DiagSplat · train-only no-reference mask construction

Fountain:强视觉语义框 → SAM 中间结果

每张 train 图独立查看;强视觉判断负责列出 transient 人与随身物,SAM 2.1 只负责像素轮廓。Raw residual 仅是辅助,低 residual 不能否定 transient。17 张 test 完全未参与。

141 transient masks · 27 clean routes · 下一步为逐帧边界安全带 critic。

IMG_6393

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6393",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6393.png",
  "routing_mask_path": "masks/IMG_6393_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6394

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6394",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6394.png",
  "routing_mask_path": "masks/IMG_6394_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6395

transient

transient_person · 1 semantic box(es) · raw 1.20% · route 2.10%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6395",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6395",
    "boxes_norm_xyxy": [
      [
        0.9008,
        0.5132,
        1,
        0.8492
      ]
    ],
    "description": "画面右缘仅部分可见的人及黑色手提袋",
    "confidence": 0.82
  },
  "boxes_xyxy": [
    [
      454.0032043457031,
      193.98959350585938,
      504.0,
      320.9975891113281
    ]
  ],
  "sam_predicted_iou": [
    0.83984375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6395.png",
  "routing_mask_path": "masks/IMG_6395_routing.png",
  "raw_mask_fraction": 0.012030738221214412,
  "routing_mask_fraction": 0.020964558662971363
}

IMG_6396

transient

transient_person · 1 semantic box(es) · raw 2.49% · route 3.45%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6396",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6396",
    "boxes_norm_xyxy": [
      [
        0.8869,
        0.2381,
        1,
        0.7169
      ]
    ],
    "description": "右缘行人及其黑色手提袋,身体被画面裁切",
    "confidence": 0.94
  },
  "boxes_xyxy": [
    [
      446.9975891113281,
      90.00180053710938,
      504.0,
      270.9881896972656
    ]
  ],
  "sam_predicted_iou": [
    0.90234375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6396.png",
  "routing_mask_path": "masks/IMG_6396_routing.png",
  "raw_mask_fraction": 0.02493281263122533,
  "routing_mask_fraction": 0.034549004787100025
}

IMG_6397

transient

transient_person · 1 semantic box(es) · raw 2.41% · route 3.16%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6397",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6397",
    "boxes_norm_xyxy": [
      [
        0.871,
        0.3466,
        1,
        0.8122
      ]
    ],
    "description": "右缘被裁切的行人及黑色手提袋",
    "confidence": 0.88
  },
  "boxes_xyxy": [
    [
      438.9840087890625,
      131.01480102539062,
      504.0,
      307.0115966796875
    ]
  ],
  "sam_predicted_iou": [
    0.87890625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6397.png",
  "routing_mask_path": "masks/IMG_6397_routing.png",
  "raw_mask_fraction": 0.024056227429243304,
  "routing_mask_fraction": 0.031588561350466116
}

IMG_6398

transient

transient_person · 1 semantic box(es) · raw 19.05% · route 22.12%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6398",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6398",
    "boxes_norm_xyxy": [
      [
        0.1508,
        0.0714,
        0.5119,
        1
      ]
    ],
    "description": "近景行人,含大号黑色背包与手中黑袋",
    "confidence": 0.99
  },
  "boxes_xyxy": [
    [
      76.0031967163086,
      26.989200592041016,
      257.9975891113281,
      378.0
    ]
  ],
  "sam_predicted_iou": [
    0.953125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6398.png",
  "routing_mask_path": "masks/IMG_6398_routing.png",
  "raw_mask_fraction": 0.1905076845553036,
  "routing_mask_fraction": 0.221172419585118
}

IMG_6399

transient

transient_person · 1 semantic box(es) · raw 4.92% · route 6.48%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6399",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6399",
    "boxes_norm_xyxy": [
      [
        0,
        0.1455,
        0.2381,
        0.7407
      ]
    ],
    "description": "左缘有清晰近景行人,含背包及手提袋。",
    "confidence": 0.9
  },
  "boxes_xyxy": [
    [
      0.0,
      54.999000549316406,
      120.00240325927734,
      279.9845886230469
    ]
  ],
  "sam_predicted_iou": [
    0.875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6399.png",
  "routing_mask_path": "masks/IMG_6399_routing.png",
  "raw_mask_fraction": 0.049162257495590826,
  "routing_mask_fraction": 0.06483056185437137
}

IMG_6400

transient

transient_person · 1 semantic box(es) · raw 0.70% · route 2.01%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6400",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6400",
    "boxes_norm_xyxy": [
      [
        0.5417,
        0.1481,
        0.6627,
        0.5026
      ]
    ],
    "description": "雕塑后方有被遮挡但清晰可辨的行人及背包。",
    "confidence": 0.9
  },
  "boxes_xyxy": [
    [
      273.0168151855469,
      55.9818000793457,
      334.00079345703125,
      189.98280334472656
    ]
  ],
  "sam_predicted_iou": [
    0.59375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6400.png",
  "routing_mask_path": "masks/IMG_6400_routing.png",
  "raw_mask_fraction": 0.00701793062904174,
  "routing_mask_fraction": 0.02010896951373142
}

IMG_6401

transient

transient_person · 1 semantic box(es) · raw 0.14% · route 1.30%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6401",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6401",
    "boxes_norm_xyxy": [
      [
        0.7956,
        0.2513,
        0.9544,
        0.5979
      ]
    ],
    "description": "喷泉后方行人,含背包与手提袋",
    "confidence": 0.96
  },
  "boxes_xyxy": [
    [
      400.9823913574219,
      94.99140167236328,
      481.0176086425781,
      226.00619506835938
    ]
  ],
  "sam_predicted_iou": [
    0.54296875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6401.png",
  "routing_mask_path": "masks/IMG_6401_routing.png",
  "raw_mask_fraction": 0.0014119845469051818,
  "routing_mask_fraction": 0.013017552700092383
}

IMG_6402

transient

transient_person · 1 semantic box(es) · raw 4.09% · route 5.85%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6402",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6402",
    "boxes_norm_xyxy": [
      [
        0.7976,
        0.2963,
        1,
        1
      ]
    ],
    "description": "右缘被裁切的行人及黑色手提袋",
    "confidence": 0.94
  },
  "boxes_xyxy": [
    [
      401.9903869628906,
      112.00140380859375,
      504.0,
      378.0
    ]
  ],
  "sam_predicted_iou": [
    0.953125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6402.png",
  "routing_mask_path": "masks/IMG_6402_routing.png",
  "raw_mask_fraction": 0.040853069622910895,
  "routing_mask_fraction": 0.05846875787351978
}

IMG_6403

transient

transient_person · 1 semantic box(es) · raw 17.19% · route 18.85%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6403",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6403",
    "boxes_norm_xyxy": [
      [
        0,
        0.0529,
        0.2718,
        1
      ]
    ],
    "description": "左侧近景行人,含大号背包与手提袋",
    "confidence": 0.99
  },
  "boxes_xyxy": [
    [
      0.0,
      19.996200561523438,
      136.98719787597656,
      378.0
    ]
  ],
  "sam_predicted_iou": [
    0.9765625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6403.png",
  "routing_mask_path": "masks/IMG_6403_routing.png",
  "raw_mask_fraction": 0.17188418577307465,
  "routing_mask_fraction": 0.18850781053162005
}

IMG_6404

transient

transient_person · 1 semantic box(es) · raw 4.61% · route 6.45%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6404",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6404",
    "boxes_norm_xyxy": [
      [
        0.0179,
        0.2646,
        0.2242,
        0.8095
      ]
    ],
    "description": "左侧完整行人,含背包与手提袋",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      9.021599769592285,
      100.018798828125,
      112.9968032836914,
      305.9909973144531
    ]
  ],
  "sam_predicted_iou": [
    0.921875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6404.png",
  "routing_mask_path": "masks/IMG_6404_routing.png",
  "raw_mask_fraction": 0.046060090702947844,
  "routing_mask_fraction": 0.06452611908961116
}

IMG_6405

transient

transient_person · 1 semantic box(es) · raw 1.04% · route 1.61%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6405",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6405",
    "boxes_norm_xyxy": [
      [
        0.4206,
        0.328,
        0.5357,
        0.5503
      ]
    ],
    "description": "喷泉后方被遮挡的行人及背包",
    "confidence": 0.94
  },
  "boxes_xyxy": [
    [
      211.98240661621094,
      123.98400115966797,
      269.9927978515625,
      208.01339721679688
    ]
  ],
  "sam_predicted_iou": [
    0.8125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6405.png",
  "routing_mask_path": "masks/IMG_6405_routing.png",
  "raw_mask_fraction": 0.010408793146888385,
  "routing_mask_fraction": 0.01605148232132359
}

IMG_6406

transient

transient_person · 1 semantic box(es) · raw 0.30% · route 0.76%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6406",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6406",
    "boxes_norm_xyxy": [
      [
        0.2361,
        0.2619,
        0.3591,
        0.5079
      ]
    ],
    "description": "喷泉后方被遮挡的行人及背包",
    "confidence": 0.94
  },
  "boxes_xyxy": [
    [
      118.99440002441406,
      98.99819946289062,
      180.9864044189453,
      191.9862060546875
    ]
  ],
  "sam_predicted_iou": [
    0.68359375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6406.png",
  "routing_mask_path": "masks/IMG_6406_routing.png",
  "raw_mask_fraction": 0.0030286806080456873,
  "routing_mask_fraction": 0.0075953220794490635
}

IMG_6407

transient

transient_person · 2 semantic box(es) · raw 3.12% · route 4.55%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6407",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6407",
    "boxes_norm_xyxy": [
      [
        0,
        0.2328,
        0.127,
        0.664
      ],
      [
        0.0853,
        0.2169,
        0.1468,
        0.3545
      ]
    ],
    "description": "左缘主要行人及其背包,另含后方较小行人",
    "confidence": 0.86
  },
  "boxes_xyxy": [
    [
      0.0,
      87.99839782714844,
      64.00800323486328,
      250.99200439453125
    ],
    [
      42.9911994934082,
      81.98819732666016,
      73.98719787597656,
      134.00100708007812
    ]
  ],
  "sam_predicted_iou": [
    0.9375,
    0.765625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6407.png",
  "routing_mask_path": "masks/IMG_6407_routing.png",
  "raw_mask_fraction": 0.031189636348366506,
  "routing_mask_fraction": 0.04549319727891157
}

IMG_6408

transient

transient_person · 2 semantic box(es) · raw 3.11% · route 4.67%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6408",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6408",
    "boxes_norm_xyxy": [
      [
        0.129,
        0.1772,
        0.2877,
        0.6296
      ],
      [
        0,
        0.1931,
        0.0218,
        0.3069
      ]
    ],
    "description": "左侧主要行人及手提袋;最左缘另有被裁切远景行人",
    "confidence": 0.88
  },
  "boxes_xyxy": [
    [
      65.01599884033203,
      66.98159790039062,
      145.00079345703125,
      237.98880004882812
    ],
    [
      0.0,
      72.9917984008789,
      10.987199783325195,
      116.0082015991211
    ]
  ],
  "sam_predicted_iou": [
    0.9296875,
    0.75390625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6408.png",
  "routing_mask_path": "masks/IMG_6408_routing.png",
  "raw_mask_fraction": 0.031121399176954733,
  "routing_mask_fraction": 0.04670047031158142
}

IMG_6409

transient

transient_person · 2 semantic box(es) · raw 3.23% · route 5.25%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6409",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6409",
    "boxes_norm_xyxy": [
      [
        0.2004,
        0.1693,
        0.4067,
        0.6402
      ],
      [
        0.3968,
        0.2037,
        0.4345,
        0.3016
      ]
    ],
    "description": "主要行人及黑袋,另含远景小行人",
    "confidence": 0.95
  },
  "boxes_xyxy": [
    [
      101.00160217285156,
      63.995399475097656,
      204.976806640625,
      241.99560546875
    ],
    [
      199.98719787597656,
      76.99859619140625,
      218.98800659179688,
      114.00479888916016
    ]
  ],
  "sam_predicted_iou": [
    0.91015625,
    0.87890625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6409.png",
  "routing_mask_path": "masks/IMG_6409_routing.png",
  "raw_mask_fraction": 0.03226043503821282,
  "routing_mask_fraction": 0.052495380868396745
}

IMG_6410

transient

transient_person · 2 semantic box(es) · raw 4.06% · route 5.99%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6410",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6410",
    "boxes_norm_xyxy": [
      [
        0.0119,
        0.1984,
        0.2103,
        0.7407
      ],
      [
        0,
        0.2698,
        0.0258,
        0.373
      ]
    ],
    "description": "左侧主要行人及黑袋;左缘另有被裁切远景行人",
    "confidence": 0.9
  },
  "boxes_xyxy": [
    [
      5.997600078582764,
      74.99520111083984,
      105.99120330810547,
      279.9845886230469
    ],
    [
      0.0,
      101.9843978881836,
      13.003199577331543,
      140.99400329589844
    ]
  ],
  "sam_predicted_iou": [
    0.9140625,
    0.81640625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6410.png",
  "routing_mask_path": "masks/IMG_6410_routing.png",
  "raw_mask_fraction": 0.04055387587133619,
  "routing_mask_fraction": 0.05988074242042496
}

IMG_6411

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6411",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6411.png",
  "routing_mask_path": "masks/IMG_6411_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6412

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6412",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6412.png",
  "routing_mask_path": "masks/IMG_6412_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6413

transient

transient_person · 1 semantic box(es) · raw 3.53% · route 5.11%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6413",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6413",
    "boxes_norm_xyxy": [
      [
        0.7044,
        0.2222,
        0.8968,
        0.7381
      ]
    ],
    "description": "右侧行人,含大号背包与黑色手提袋",
    "confidence": 0.99
  },
  "boxes_xyxy": [
    [
      355.0176086425781,
      83.9916000366211,
      451.9872131347656,
      279.0018005371094
    ]
  ],
  "sam_predicted_iou": [
    0.9296875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6413.png",
  "routing_mask_path": "masks/IMG_6413_routing.png",
  "raw_mask_fraction": 0.035257621567145375,
  "routing_mask_fraction": 0.05113063744016125
}

IMG_6414

transient

transient_person · 1 semantic box(es) · raw 1.73% · route 2.65%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6414",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6414",
    "boxes_norm_xyxy": [
      [
        0.5516,
        0.4524,
        0.6548,
        0.8122
      ]
    ],
    "description": "喷泉后方行人及背包",
    "confidence": 0.95
  },
  "boxes_xyxy": [
    [
      278.00640869140625,
      171.0072021484375,
      330.0191955566406,
      307.0115966796875
    ]
  ],
  "sam_predicted_iou": [
    0.9296875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6414.png",
  "routing_mask_path": "masks/IMG_6414_routing.png",
  "raw_mask_fraction": 0.01731124548584866,
  "routing_mask_fraction": 0.02647602250776854
}

IMG_6415

transient

transient_person · 1 semantic box(es) · raw 0.26% · route 1.35%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6415",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6415",
    "boxes_norm_xyxy": [
      [
        0.7381,
        0.127,
        0.8452,
        0.3968
      ]
    ],
    "description": "右上方被喷泉遮挡的行人及背包",
    "confidence": 0.87
  },
  "boxes_xyxy": [
    [
      372.0024108886719,
      48.00600051879883,
      425.9808044433594,
      149.9904022216797
    ]
  ],
  "sam_predicted_iou": [
    0.76953125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6415.png",
  "routing_mask_path": "masks/IMG_6415_routing.png",
  "raw_mask_fraction": 0.002551020408163265,
  "routing_mask_fraction": 0.013489963886789283
}

IMG_6416

transient

transient_person · 1 semantic box(es) · raw 0.82% · route 1.36%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6416",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6416",
    "boxes_norm_xyxy": [
      [
        0.6706,
        0.0714,
        0.7778,
        0.254
      ]
    ],
    "description": "喷泉后方远景行人及背包",
    "confidence": 0.86
  },
  "boxes_xyxy": [
    [
      337.9823913574219,
      26.989200592041016,
      392.0111999511719,
      96.01200103759766
    ]
  ],
  "sam_predicted_iou": [
    0.8046875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6416.png",
  "routing_mask_path": "masks/IMG_6416_routing.png",
  "raw_mask_fraction": 0.008246199714453683,
  "routing_mask_fraction": 0.013579197110943143
}

IMG_6417

transient

transient_person · 2 semantic box(es) · raw 1.33% · route 2.32%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6417",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6417",
    "boxes_norm_xyxy": [
      [
        0.4365,
        0.4127,
        0.5417,
        0.6111
      ],
      [
        0.881,
        0.4048,
        0.9286,
        0.5026
      ]
    ],
    "description": "喷泉后方主要行人及手提袋;右侧棚下另有小行人",
    "confidence": 0.86
  },
  "boxes_xyxy": [
    [
      219.99600219726562,
      156.00059509277344,
      273.0168151855469,
      230.9958038330078
    ],
    [
      444.02398681640625,
      153.014404296875,
      468.014404296875,
      189.98280334472656
    ]
  ],
  "sam_predicted_iou": [
    0.90234375,
    0.61328125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6417.png",
  "routing_mask_path": "masks/IMG_6417_routing.png",
  "raw_mask_fraction": 0.013264256319811875,
  "routing_mask_fraction": 0.02324787939867305
}

IMG_6418

transient

transient_person · 2 semantic box(es) · raw 1.59% · route 2.65%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6418",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6418",
    "boxes_norm_xyxy": [
      [
        0.3373,
        0.5582,
        0.4563,
        0.8492
      ],
      [
        0.881,
        0.4762,
        0.9365,
        0.5608
      ]
    ],
    "description": "树旁主要行人及黑袋;右侧棚下另有小行人",
    "confidence": 0.85
  },
  "boxes_xyxy": [
    [
      169.99920654296875,
      210.99960327148438,
      229.97520446777344,
      320.9975891113281
    ],
    [
      444.02398681640625,
      180.00360107421875,
      471.9960021972656,
      211.98240661621094
    ]
  ],
  "sam_predicted_iou": [
    0.89453125,
    0.69140625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6418.png",
  "routing_mask_path": "masks/IMG_6418_routing.png",
  "raw_mask_fraction": 0.015899260938943477,
  "routing_mask_fraction": 0.02651801461325271
}

IMG_6419

transient

transient_person · 1 semantic box(es) · raw 1.98% · route 3.35%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6419",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6419",
    "boxes_norm_xyxy": [
      [
        0.1369,
        0.0741,
        0.2877,
        0.4392
      ]
    ],
    "description": "左上方行人及黑色手提袋",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      68.99759674072266,
      28.00979995727539,
      145.00079345703125,
      166.01759338378906
    ]
  ],
  "sam_predicted_iou": [
    0.92578125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6419.png",
  "routing_mask_path": "masks/IMG_6419_routing.png",
  "raw_mask_fraction": 0.019815024775342235,
  "routing_mask_fraction": 0.033483455110439236
}

IMG_6420

transient

transient_person · 1 semantic box(es) · raw 2.00% · route 3.41%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6420",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6420",
    "boxes_norm_xyxy": [
      [
        0.1964,
        0.0635,
        0.3452,
        0.4524
      ]
    ],
    "description": "左侧行人及黑色手提袋",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      98.98560333251953,
      24.003000259399414,
      173.98080444335938,
      171.0072021484375
    ]
  ],
  "sam_predicted_iou": [
    0.921875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6420.png",
  "routing_mask_path": "masks/IMG_6420_routing.png",
  "raw_mask_fraction": 0.019956748131351305,
  "routing_mask_fraction": 0.034071344587217604
}

IMG_6421

transient

transient_person · 1 semantic box(es) · raw 1.43% · route 2.96%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6421",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6421",
    "boxes_norm_xyxy": [
      [
        0.2361,
        0.4127,
        0.369,
        0.7751
      ]
    ],
    "description": "左侧行人,含背包与手提袋",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      118.99440002441406,
      156.00059509277344,
      185.9759979248047,
      292.98779296875
    ]
  ],
  "sam_predicted_iou": [
    0.828125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6421.png",
  "routing_mask_path": "masks/IMG_6421_routing.png",
  "raw_mask_fraction": 0.014282564877802974,
  "routing_mask_fraction": 0.02962018140589569
}

IMG_6422

transient

transient_person · 2 semantic box(es) · raw 1.88% · route 3.45%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6422",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6422",
    "boxes_norm_xyxy": [
      [
        0.2103,
        0.5026,
        0.3254,
        0.7434
      ],
      [
        0.0972,
        0.5132,
        0.1627,
        0.7354
      ]
    ],
    "description": "喷泉旁主要行人及背包,另含左侧第二名行人",
    "confidence": 0.96
  },
  "boxes_xyxy": [
    [
      105.99120330810547,
      189.98280334472656,
      164.00160217285156,
      281.00518798828125
    ],
    [
      48.988800048828125,
      193.98959350585938,
      82.00080108642578,
      277.981201171875
    ]
  ],
  "sam_predicted_iou": [
    0.9140625,
    0.85546875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6422.png",
  "routing_mask_path": "masks/IMG_6422_routing.png",
  "raw_mask_fraction": 0.018843957336020827,
  "routing_mask_fraction": 0.03452275972117242
}

IMG_6423

transient

transient_person · 2 semantic box(es) · raw 2.22% · route 3.80%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6423",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6423",
    "boxes_norm_xyxy": [
      [
        0.0992,
        0.6111,
        0.2341,
        1
      ],
      [
        0.3571,
        0.6429,
        0.4127,
        0.8598
      ]
    ],
    "description": "左侧被画面裁切的主要行人及背包/黑袋,另含树后第二人",
    "confidence": 0.92
  },
  "boxes_xyxy": [
    [
      49.99679946899414,
      230.9958038330078,
      117.98639678955078,
      378.0
    ],
    [
      179.9783935546875,
      243.01620483398438,
      208.00079345703125,
      325.00439453125
    ]
  ],
  "sam_predicted_iou": [
    0.90625,
    0.71875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6423.png",
  "routing_mask_path": "masks/IMG_6423_routing.png",
  "raw_mask_fraction": 0.022187578735197784,
  "routing_mask_fraction": 0.03803434954228605
}

IMG_6424

transient

transient_person · 1 semantic box(es) · raw 2.76% · route 8.64%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6424",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6424",
    "boxes_norm_xyxy": [
      [
        0.2738,
        0.1746,
        0.4127,
        0.5688
      ]
    ],
    "description": "左侧有清晰行人,含背包和手提袋。",
    "confidence": 0.9
  },
  "boxes_xyxy": [
    [
      137.9951934814453,
      65.9988021850586,
      208.00079345703125,
      215.0063934326172
    ]
  ],
  "sam_predicted_iou": [
    0.56640625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6424.png",
  "routing_mask_path": "masks/IMG_6424_routing.png",
  "raw_mask_fraction": 0.027578315276727976,
  "routing_mask_fraction": 0.0863567649281935
}

IMG_6425

transient

transient_person · 1 semantic box(es) · raw 1.07% · route 1.70%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6425",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6425",
    "boxes_norm_xyxy": [
      [
        0.3929,
        0.336,
        0.5079,
        0.5899
      ]
    ],
    "description": "喷泉后方部分遮挡的行人及背包",
    "confidence": 0.91
  },
  "boxes_xyxy": [
    [
      198.0216064453125,
      127.00800323486328,
      255.98159790039062,
      222.98219299316406
    ]
  ],
  "sam_predicted_iou": [
    0.90625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6425.png",
  "routing_mask_path": "masks/IMG_6425_routing.png",
  "raw_mask_fraction": 0.010702737885277569,
  "routing_mask_fraction": 0.01699630469471739
}

IMG_6426

transient

transient_person · 1 semantic box(es) · raw 0.86% · route 1.58%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6426",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6426",
    "boxes_norm_xyxy": [
      [
        0.5437,
        0.5688,
        0.6429,
        0.7857
      ]
    ],
    "description": "雕塑后方部分遮挡的行人及背包",
    "confidence": 0.89
  },
  "boxes_xyxy": [
    [
      274.0248107910156,
      215.0063934326172,
      324.0216064453125,
      296.9945983886719
    ]
  ],
  "sam_predicted_iou": [
    0.87109375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6426.png",
  "routing_mask_path": "masks/IMG_6426_routing.png",
  "raw_mask_fraction": 0.00860838162425464,
  "routing_mask_fraction": 0.01579952968841858
}

IMG_6427

transient

transient_person · 1 semantic box(es) · raw 0.21% · route 0.55%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6427",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6427",
    "boxes_norm_xyxy": [
      [
        0.8929,
        0.6032,
        0.9345,
        0.7249
      ]
    ],
    "description": "右侧车辆旁远景小行人",
    "confidence": 0.63
  },
  "boxes_xyxy": [
    [
      450.0216064453125,
      228.0095977783203,
      470.9880065917969,
      274.01220703125
    ]
  ],
  "sam_predicted_iou": [
    0.7734375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6427.png",
  "routing_mask_path": "masks/IMG_6427_routing.png",
  "raw_mask_fraction": 0.002062862181909801,
  "routing_mask_fraction": 0.0054589737129419665
}

IMG_6428

transient

transient_person · 2 semantic box(es) · raw 2.60% · route 4.25%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6428",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6428",
    "boxes_norm_xyxy": [
      [
        0.8095,
        0.1032,
        0.9603,
        0.4683
      ],
      [
        0.748,
        0.1111,
        0.7937,
        0.2566
      ]
    ],
    "description": "右侧主要行人及手提袋,另含其后方小行人",
    "confidence": 0.93
  },
  "boxes_xyxy": [
    [
      407.9880065917969,
      39.00960159301758,
      483.9912109375,
      177.01739501953125
    ],
    [
      376.99200439453125,
      41.99580001831055,
      400.0248107910156,
      96.99479675292969
    ]
  ],
  "sam_predicted_iou": [
    0.93359375,
    0.8125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6428.png",
  "routing_mask_path": "masks/IMG_6428_routing.png",
  "raw_mask_fraction": 0.02600886033425716,
  "routing_mask_fraction": 0.04253275384227765
}

IMG_6429

transient

transient_person · 2 semantic box(es) · raw 3.21% · route 5.26%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6429",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6429",
    "boxes_norm_xyxy": [
      [
        0.8135,
        0.1455,
        0.9802,
        0.5529
      ],
      [
        0.0734,
        0.1825,
        0.1548,
        0.5159
      ]
    ],
    "description": "右侧主要行人及背包/黑袋,另含左侧第二人",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      410.0039978027344,
      54.999000549316406,
      494.02081298828125,
      208.99620056152344
    ],
    [
      36.99359893798828,
      68.98500061035156,
      78.01920318603516,
      195.01019287109375
    ]
  ],
  "sam_predicted_iou": [
    0.92578125,
    0.8515625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6429.png",
  "routing_mask_path": "masks/IMG_6429_routing.png",
  "raw_mask_fraction": 0.0320819685899051,
  "routing_mask_fraction": 0.05261610817166373
}

IMG_6430

transient

transient_person · 1 semantic box(es) · raw 0.58% · route 1.11%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6430",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6430",
    "boxes_norm_xyxy": [
      [
        0.2262,
        0.4021,
        0.3036,
        0.6164
      ]
    ],
    "description": "树旁远景行人及背包",
    "confidence": 0.88
  },
  "boxes_xyxy": [
    [
      114.00479888916016,
      151.99380493164062,
      153.014404296875,
      232.99920654296875
    ]
  ],
  "sam_predicted_iou": [
    0.85546875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6430.png",
  "routing_mask_path": "masks/IMG_6430_routing.png",
  "raw_mask_fraction": 0.005779163517258755,
  "routing_mask_fraction": 0.011148904006046864
}

IMG_6431

transient

transient_person · 1 semantic box(es) · raw 0.66% · route 1.30%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6431",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6431",
    "boxes_norm_xyxy": [
      [
        0.4246,
        0.5026,
        0.5298,
        0.7566
      ]
    ],
    "description": "树与雕塑之间部分遮挡的行人及随身袋",
    "confidence": 0.86
  },
  "boxes_xyxy": [
    [
      213.99839782714844,
      189.98280334472656,
      267.0191955566406,
      285.99481201171875
    ]
  ],
  "sam_predicted_iou": [
    0.89453125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6431.png",
  "routing_mask_path": "masks/IMG_6431_routing.png",
  "raw_mask_fraction": 0.006619005626942135,
  "routing_mask_fraction": 0.013017552700092383
}

IMG_6432

transient

transient_person · 3 semantic box(es) · raw 2.53% · route 3.95%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6432",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6432",
    "boxes_norm_xyxy": [
      [
        0,
        0.3519,
        0.0933,
        0.6905
      ],
      [
        0.5615,
        0.3386,
        0.619,
        0.5265
      ],
      [
        0.6468,
        0.336,
        0.7024,
        0.5397
      ]
    ],
    "description": "左缘被裁切的行人及背包,另含道路旁两名远景行人",
    "confidence": 0.92
  },
  "boxes_xyxy": [
    [
      0.0,
      133.01820373535156,
      47.02320098876953,
      261.0090026855469
    ],
    [
      282.9960021972656,
      127.99079895019531,
      311.97601318359375,
      199.01699829101562
    ],
    [
      325.9872131347656,
      127.00800323486328,
      354.0096130371094,
      204.00660705566406
    ]
  ],
  "sam_predicted_iou": [
    0.91796875,
    0.78515625,
    0.84765625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6432.png",
  "routing_mask_path": "masks/IMG_6432_routing.png",
  "raw_mask_fraction": 0.02530549256739733,
  "routing_mask_fraction": 0.0394883261946754
}

IMG_6433

transient

transient_person · 2 semantic box(es) · raw 2.80% · route 4.32%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6433",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6433",
    "boxes_norm_xyxy": [
      [
        0.123,
        0.2857,
        0.2619,
        0.6323
      ],
      [
        0.8294,
        0.3175,
        0.9147,
        0.5476
      ]
    ],
    "description": "左侧主要行人及背包/手提袋,另含右侧第二人",
    "confidence": 0.97
  },
  "boxes_xyxy": [
    [
      61.992000579833984,
      107.99459838867188,
      131.9976043701172,
      239.0093994140625
    ],
    [
      418.0176086425781,
      120.01499938964844,
      461.0087890625,
      206.9927978515625
    ]
  ],
  "sam_predicted_iou": [
    0.91796875,
    0.84765625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6433.png",
  "routing_mask_path": "masks/IMG_6433_routing.png",
  "raw_mask_fraction": 0.02802448139749727,
  "routing_mask_fraction": 0.04319412950365331
}

IMG_6434

transient

transient_person · 2 semantic box(es) · raw 3.73% · route 5.81%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6434",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6434",
    "boxes_norm_xyxy": [
      [
        0.6171,
        0.2222,
        0.7698,
        0.6085
      ],
      [
        0.9067,
        0.2831,
        0.9782,
        0.5238
      ]
    ],
    "description": "树前主要行人及背包/手提袋,另含右侧第二人",
    "confidence": 0.97
  },
  "boxes_xyxy": [
    [
      311.0184020996094,
      83.9916000366211,
      387.97918701171875,
      230.01300048828125
    ],
    [
      456.976806640625,
      107.01180267333984,
      493.0127868652344,
      197.99639892578125
    ]
  ],
  "sam_predicted_iou": [
    0.8515625,
    0.890625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6434.png",
  "routing_mask_path": "masks/IMG_6434_routing.png",
  "raw_mask_fraction": 0.03731523473586966,
  "routing_mask_fraction": 0.05806458385823465
}

IMG_6435

transient

transient_person · 1 semantic box(es) · raw 0.80% · route 1.55%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6435",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6435",
    "boxes_norm_xyxy": [
      [
        0.9345,
        0.1905,
        1,
        0.5582
      ]
    ],
    "description": "右缘有被画面裁切但清晰可辨的行人。",
    "confidence": 0.9
  },
  "boxes_xyxy": [
    [
      470.9880065917969,
      72.00900268554688,
      504.0,
      210.99960327148438
    ]
  ],
  "sam_predicted_iou": [
    0.6328125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6435.png",
  "routing_mask_path": "masks/IMG_6435_routing.png",
  "raw_mask_fraction": 0.008025741160661796,
  "routing_mask_fraction": 0.015453094818174183
}

IMG_6437

transient

transient_person · 2 semantic box(es) · raw 2.94% · route 4.59%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6437",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6437",
    "boxes_norm_xyxy": [
      [
        0.754,
        0,
        0.9246,
        0.4101
      ],
      [
        0.4306,
        0.0503,
        0.5238,
        0.2566
      ]
    ],
    "description": "右上方被裁切的主要行人及背包,另含道路旁第二人",
    "confidence": 0.93
  },
  "boxes_xyxy": [
    [
      380.0159912109375,
      0.0,
      465.9984130859375,
      155.01780700683594
    ],
    [
      217.02239990234375,
      19.013399124145508,
      263.9952087402344,
      96.99479675292969
    ]
  ],
  "sam_predicted_iou": [
    0.86328125,
    0.8515625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6437.png",
  "routing_mask_path": "masks/IMG_6437_routing.png",
  "raw_mask_fraction": 0.02943121693121693,
  "routing_mask_fraction": 0.045949861426051904
}

IMG_6438

transient

transient_person · 2 semantic box(es) · raw 1.61% · route 2.43%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6438",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6438",
    "boxes_norm_xyxy": [
      [
        0.5595,
        0,
        0.6845,
        0.1878
      ],
      [
        0.2738,
        0,
        0.3353,
        0.127
      ]
    ],
    "description": "上缘被裁切的主要行人及背包,另含左上远景行人",
    "confidence": 0.87
  },
  "boxes_xyxy": [
    [
      281.9880065917969,
      0.0,
      344.9880065917969,
      70.9884033203125
    ],
    [
      137.9951934814453,
      0.0,
      168.99119567871094,
      48.00600051879883
    ]
  ],
  "sam_predicted_iou": [
    0.88671875,
    0.87890625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6438.png",
  "routing_mask_path": "masks/IMG_6438_routing.png",
  "raw_mask_fraction": 0.016061980347694634,
  "routing_mask_fraction": 0.024250440917107582
}

IMG_6439

transient

transient_person · 1 semantic box(es) · raw 1.51% · route 2.27%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6439",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6439",
    "boxes_norm_xyxy": [
      [
        0.502,
        0.3228,
        0.6151,
        0.6111
      ]
    ],
    "description": "喷泉后方行人及黑色手提袋",
    "confidence": 0.97
  },
  "boxes_xyxy": [
    [
      253.00799560546875,
      122.01840209960938,
      310.0104064941406,
      230.9958038330078
    ]
  ],
  "sam_predicted_iou": [
    0.91796875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6439.png",
  "routing_mask_path": "masks/IMG_6439_routing.png",
  "raw_mask_fraction": 0.015122406987486353,
  "routing_mask_fraction": 0.02272822709330646
}

IMG_6440

transient

transient_person · 1 semantic box(es) · raw 1.63% · route 2.40%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6440",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6440",
    "boxes_norm_xyxy": [
      [
        0.4484,
        0.4365,
        0.5575,
        0.7407
      ]
    ],
    "description": "喷泉后方行人及背包/手提袋",
    "confidence": 0.96
  },
  "boxes_xyxy": [
    [
      225.9936065673828,
      164.9969940185547,
      280.9800109863281,
      279.9845886230469
    ]
  ],
  "sam_predicted_iou": [
    0.8984375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6440.png",
  "routing_mask_path": "masks/IMG_6440_routing.png",
  "raw_mask_fraction": 0.01628768791467204,
  "routing_mask_fraction": 0.023987990257831528
}

IMG_6441

transient

transient_person · 1 semantic box(es) · raw 1.89% · route 2.78%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6441",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6441",
    "boxes_norm_xyxy": [
      [
        0.5774,
        0.4101,
        0.7202,
        0.7275
      ]
    ],
    "description": "喷泉后方行人及黑色手提袋",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      291.0096130371094,
      155.01780700683594,
      362.9808044433594,
      274.9949951171875
    ]
  ],
  "sam_predicted_iou": [
    0.921875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6441.png",
  "routing_mask_path": "masks/IMG_6441_routing.png",
  "raw_mask_fraction": 0.01890169648106156,
  "routing_mask_fraction": 0.027814520870076426
}

IMG_6442

transient

transient_person · 1 semantic box(es) · raw 2.00% · route 2.89%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6442",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6442",
    "boxes_norm_xyxy": [
      [
        0.6488,
        0.1164,
        0.7758,
        0.4233
      ]
    ],
    "description": "喷泉后方有清晰行人及手提袋。",
    "confidence": 0.9
  },
  "boxes_xyxy": [
    [
      326.9952087402344,
      43.99919891357422,
      391.0032043457031,
      160.0074005126953
    ]
  ],
  "sam_predicted_iou": [
    0.91015625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6442.png",
  "routing_mask_path": "masks/IMG_6442_routing.png",
  "raw_mask_fraction": 0.0200302343159486,
  "routing_mask_fraction": 0.028895817586293778
}

IMG_6443

transient

transient_person · 1 semantic box(es) · raw 2.95% · route 3.82%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6443",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6443",
    "boxes_norm_xyxy": [
      [
        0.7758,
        0,
        0.9643,
        0.2672
      ]
    ],
    "description": "右上缘被裁切的行人及手提袋",
    "confidence": 0.9
  },
  "boxes_xyxy": [
    [
      391.0032043457031,
      0.0,
      486.0072021484375,
      101.00160217285156
    ]
  ],
  "sam_predicted_iou": [
    0.9375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6443.png",
  "routing_mask_path": "masks/IMG_6443_routing.png",
  "raw_mask_fraction": 0.02953094818174183,
  "routing_mask_fraction": 0.03818132191148064
}

IMG_6444

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6444",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6444.png",
  "routing_mask_path": "masks/IMG_6444_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6445

transient

transient_person · 1 semantic box(es) · raw 0.57% · route 1.12%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6445",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6445",
    "boxes_norm_xyxy": [
      [
        0.5238,
        0.1508,
        0.6171,
        0.3862
      ]
    ],
    "description": "道路对面远景行人",
    "confidence": 0.88
  },
  "boxes_xyxy": [
    [
      263.9952087402344,
      57.00239944458008,
      311.0184020996094,
      145.9835968017578
    ]
  ],
  "sam_predicted_iou": [
    0.8515625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6445.png",
  "routing_mask_path": "masks/IMG_6445_routing.png",
  "raw_mask_fraction": 0.005700428319475938,
  "routing_mask_fraction": 0.011222390190644159
}

IMG_6446

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6446",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6446.png",
  "routing_mask_path": "masks/IMG_6446_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6447

transient

transient_person · 1 semantic box(es) · raw 1.69% · route 2.77%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6447",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6447",
    "boxes_norm_xyxy": [
      [
        0.9365,
        0.1825,
        1,
        0.7989
      ]
    ],
    "description": "右缘被严重裁切的行人及黑色随身袋",
    "confidence": 0.81
  },
  "boxes_xyxy": [
    [
      471.9960021972656,
      68.98500061035156,
      504.0,
      301.98419189453125
    ]
  ],
  "sam_predicted_iou": [
    0.91796875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6447.png",
  "routing_mask_path": "masks/IMG_6447_routing.png",
  "raw_mask_fraction": 0.01693856554967666,
  "routing_mask_fraction": 0.02774628369866465
}

IMG_6448

transient

transient_person · 1 semantic box(es) · raw 2.51% · route 3.83%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6448",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6448",
    "boxes_norm_xyxy": [
      [
        0.756,
        0.1429,
        0.8869,
        0.5767
      ]
    ],
    "description": "One pedestrian at right, including worn backpack.",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      381.02398681640625,
      54.01620101928711,
      446.9975891113281,
      217.9925994873047
    ]
  ],
  "sam_predicted_iou": [
    0.91796875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6448.png",
  "routing_mask_path": "masks/IMG_6448_routing.png",
  "raw_mask_fraction": 0.025106030066347526,
  "routing_mask_fraction": 0.03830204921474763
}

IMG_6449

transient

transient_person · 1 semantic box(es) · raw 3.00% · route 4.50%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6449",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6449",
    "boxes_norm_xyxy": [
      [
        0.7718,
        0.0,
        0.9385,
        0.4286
      ]
    ],
    "description": "Top-clipped pedestrian at right, including black carried/worn bag.",
    "confidence": 0.96
  },
  "boxes_xyxy": [
    [
      388.9872131347656,
      0.0,
      473.0039978027344,
      162.01080322265625
    ]
  ],
  "sam_predicted_iou": [
    0.9140625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6449.png",
  "routing_mask_path": "masks/IMG_6449_routing.png",
  "raw_mask_fraction": 0.02995086923658352,
  "routing_mask_fraction": 0.04503128411858571
}

IMG_6450

transient

transient_person · 1 semantic box(es) · raw 2.39% · route 3.08%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6450",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6450",
    "boxes_norm_xyxy": [
      [
        0.9107,
        0.0,
        1.0,
        0.4286
      ]
    ],
    "description": "Severely right-clipped pedestrian.",
    "confidence": 0.96
  },
  "boxes_xyxy": [
    [
      458.9927978515625,
      0.0,
      504.0,
      162.01080322265625
    ]
  ],
  "sam_predicted_iou": [
    0.94921875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6450.png",
  "routing_mask_path": "masks/IMG_6450_routing.png",
  "raw_mask_fraction": 0.023883009994121105,
  "routing_mask_fraction": 0.030764466280339298
}

IMG_6451

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6451",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6451.png",
  "routing_mask_path": "masks/IMG_6451_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6452

transient

transient_person · 1 semantic box(es) · raw 4.55% · route 6.29%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6452",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6452",
    "boxes_norm_xyxy": [
      [
        0.748,
        0.0,
        0.9544,
        0.4735
      ]
    ],
    "description": "Top-clipped pedestrian and backpack at right.",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      376.99200439453125,
      0.0,
      481.0176086425781,
      178.98300170898438
    ]
  ],
  "sam_predicted_iou": [
    0.91796875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6452.png",
  "routing_mask_path": "masks/IMG_6452_routing.png",
  "raw_mask_fraction": 0.04551419333165365,
  "routing_mask_fraction": 0.06291467204165617
}

IMG_6453

transient

transient_person · 1 semantic box(es) · raw 2.15% · route 3.65%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6453",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6453",
    "boxes_norm_xyxy": [
      [
        0.7044,
        0.2593,
        0.8313,
        0.6984
      ]
    ],
    "description": "One pedestrian walking away, including backpack.",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      355.0176086425781,
      98.0154037475586,
      418.9751892089844,
      263.9952087402344
    ]
  ],
  "sam_predicted_iou": [
    0.90234375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6453.png",
  "routing_mask_path": "masks/IMG_6453_routing.png",
  "raw_mask_fraction": 0.021505207021080035,
  "routing_mask_fraction": 0.03648589065255732
}

IMG_6454

transient

transient_person · 1 semantic box(es) · raw 2.12% · route 3.55%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6454",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6454",
    "boxes_norm_xyxy": [
      [
        0.6825,
        0.4312,
        0.8274,
        0.828
      ]
    ],
    "description": "One pedestrian, including black bag at left side.",
    "confidence": 0.97
  },
  "boxes_xyxy": [
    [
      343.9800109863281,
      162.9936065673828,
      417.0096130371094,
      312.9840087890625
    ]
  ],
  "sam_predicted_iou": [
    0.89453125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6454.png",
  "routing_mask_path": "masks/IMG_6454_routing.png",
  "raw_mask_fraction": 0.021158772150835642,
  "routing_mask_fraction": 0.03551482321323591
}

IMG_6455

transient

transient_person · 1 semantic box(es) · raw 5.94% · route 7.90%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6455",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6455",
    "boxes_norm_xyxy": [
      [
        0.627,
        0.0,
        0.8948,
        0.6058
      ]
    ],
    "description": "Large top-clipped pedestrian with backpack and carried black bag.",
    "confidence": 0.99
  },
  "boxes_xyxy": [
    [
      316.00799560546875,
      0.0,
      450.97918701171875,
      228.99240112304688
    ]
  ],
  "sam_predicted_iou": [
    0.9296875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6455.png",
  "routing_mask_path": "masks/IMG_6455_routing.png",
  "raw_mask_fraction": 0.05940308222054254,
  "routing_mask_fraction": 0.07901864449483498
}

IMG_6456

transient

transient_person · 1 semantic box(es) · raw 3.02% · route 5.11%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6456",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6456",
    "boxes_norm_xyxy": [
      [
        0.619,
        0.3333,
        0.7738,
        0.8228
      ]
    ],
    "description": "One pedestrian walking away, including backpack and thin carried item.",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      311.97601318359375,
      125.98739624023438,
      389.9952087402344,
      311.0184020996094
    ]
  ],
  "sam_predicted_iou": [
    0.9140625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6456.png",
  "routing_mask_path": "masks/IMG_6456_routing.png",
  "raw_mask_fraction": 0.030239564961787185,
  "routing_mask_fraction": 0.05108339632149156
}

IMG_6457

transient

transient_person · 1 semantic box(es) · raw 3.26% · route 4.88%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6457",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6457",
    "boxes_norm_xyxy": [
      [
        0.5595,
        0.5026,
        0.7341,
        0.9497
      ]
    ],
    "description": "One pedestrian facing camera, including backpack and black hand bag.",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      281.9880065917969,
      189.98280334472656,
      369.98638916015625,
      358.9866027832031
    ]
  ],
  "sam_predicted_iou": [
    0.92578125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6457.png",
  "routing_mask_path": "masks/IMG_6457_routing.png",
  "raw_mask_fraction": 0.03261211892164273,
  "routing_mask_fraction": 0.048752834467120185
}

IMG_6458

transient

transient_person · 1 semantic box(es) · raw 3.62% · route 5.01%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6458",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6458",
    "boxes_norm_xyxy": [
      [
        0.7401,
        0.5106,
        0.8909,
        1.0
      ]
    ],
    "description": "Bottom-clipped pedestrian, including backpack and bundled black bag.",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      373.0104064941406,
      193.00680541992188,
      449.01361083984375,
      378.0
    ]
  ],
  "sam_predicted_iou": [
    0.91796875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6458.png",
  "routing_mask_path": "masks/IMG_6458_routing.png",
  "raw_mask_fraction": 0.036212941966910224,
  "routing_mask_fraction": 0.05010707986898463
}

IMG_6459

transient

transient_person · 1 semantic box(es) · raw 7.62% · route 11.17%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6459",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6459",
    "boxes_norm_xyxy": [
      [
        0.633,
        0.0926,
        0.881,
        0.9233
      ]
    ],
    "description": "One prominent pedestrian with backpack and large black bag across torso.",
    "confidence": 0.99
  },
  "boxes_xyxy": [
    [
      319.0320129394531,
      35.00279998779297,
      444.02398681640625,
      349.00738525390625
    ]
  ],
  "sam_predicted_iou": [
    0.94921875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6459.png",
  "routing_mask_path": "masks/IMG_6459_routing.png",
  "raw_mask_fraction": 0.07618417737465356,
  "routing_mask_fraction": 0.11168325354833292
}

IMG_6461

transient

transient_person · 1 semantic box(es) · raw 10.89% · route 13.43%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6461",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6461",
    "boxes_norm_xyxy": [
      [
        0.6885,
        0.2222,
        0.9306,
        1.0
      ]
    ],
    "description": "Bottom-clipped older pedestrian walking away, including carried paper.",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      347.0039978027344,
      83.9916000366211,
      469.02239990234375,
      378.0
    ]
  ],
  "sam_predicted_iou": [
    0.96484375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6461.png",
  "routing_mask_path": "masks/IMG_6461_routing.png",
  "raw_mask_fraction": 0.10885403544133702,
  "routing_mask_fraction": 0.13429600235155792
}

IMG_6462

transient

transient_person · 2 semantic box(es) · raw 22.46% · route 26.33%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6462",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6462",
    "boxes_norm_xyxy": [
      [
        0.0,
        0.2381,
        0.4623,
        1.0
      ],
      [
        0.7798,
        0.3624,
        0.9087,
        0.8042
      ]
    ],
    "description": "Large foreground pedestrian with backpack plus one background older pedestrian.",
    "confidence": 0.97
  },
  "boxes_xyxy": [
    [
      0.0,
      90.00180053710938,
      232.99920654296875,
      378.0
    ],
    [
      393.0191955566406,
      136.98719787597656,
      457.98480224609375,
      303.98760986328125
    ]
  ],
  "sam_predicted_iou": [
    0.95703125,
    0.953125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6462.png",
  "routing_mask_path": "masks/IMG_6462_routing.png",
  "raw_mask_fraction": 0.22461577223481985,
  "routing_mask_fraction": 0.26328525237255396
}

IMG_6463

transient

transient_person · 1 semantic box(es) · raw 1.88% · route 2.97%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6463",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6463",
    "boxes_norm_xyxy": [
      [
        0.879,
        0.4762,
        1,
        0.8757
      ]
    ],
    "description": "右侧有清晰背向行走的人,身体部分贴近画面边缘。",
    "confidence": 0.9
  },
  "boxes_xyxy": [
    [
      443.0159912109375,
      180.00360107421875,
      504.0,
      331.01458740234375
    ]
  ],
  "sam_predicted_iou": [
    0.93359375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6463.png",
  "routing_mask_path": "masks/IMG_6463_routing.png",
  "raw_mask_fraction": 0.018754724111866967,
  "routing_mask_fraction": 0.02970416561686403
}

IMG_6464

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6464",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6464.png",
  "routing_mask_path": "masks/IMG_6464_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6465

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6465",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6465.png",
  "routing_mask_path": "masks/IMG_6465_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6466

transient

transient_person · 1 semantic box(es) · raw 1.46% · route 2.08%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6466",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6466",
    "boxes_norm_xyxy": [
      [
        0.0,
        0.0,
        0.1012,
        0.2751
      ]
    ],
    "description": "Small top-left clipped portion of pedestrian and black bag.",
    "confidence": 0.91
  },
  "boxes_xyxy": [
    [
      0.0,
      0.0,
      51.004798889160156,
      103.98780059814453
    ]
  ],
  "sam_predicted_iou": [
    0.95703125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6466.png",
  "routing_mask_path": "masks/IMG_6466_routing.png",
  "raw_mask_fraction": 0.014613252708490804,
  "routing_mask_fraction": 0.02081758629377677
}

IMG_6467

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6467",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6467.png",
  "routing_mask_path": "masks/IMG_6467_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6468

transient

transient_person · 1 semantic box(es) · raw 14.59% · route 18.38%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6468",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6468",
    "boxes_norm_xyxy": [
      [
        0.1151,
        0.082,
        0.4782,
        1.0
      ]
    ],
    "description": "Large foreground pedestrian, including backpack and carried black bag.",
    "confidence": 0.99
  },
  "boxes_xyxy": [
    [
      58.010398864746094,
      30.996000289916992,
      241.01280212402344,
      378.0
    ]
  ],
  "sam_predicted_iou": [
    0.9453125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6468.png",
  "routing_mask_path": "masks/IMG_6468_routing.png",
  "raw_mask_fraction": 0.14590157050474512,
  "routing_mask_fraction": 0.18379944570420761
}

IMG_6469

transient

transient_person · 1 semantic box(es) · raw 13.48% · route 15.38%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6469",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6469",
    "boxes_norm_xyxy": [
      [
        0.5099,
        0.3201,
        0.8373,
        1.0
      ]
    ],
    "description": "Bottom-clipped pedestrian at right, including large backpack.",
    "confidence": 0.99
  },
  "boxes_xyxy": [
    [
      256.9895935058594,
      120.997802734375,
      421.99920654296875,
      378.0
    ]
  ],
  "sam_predicted_iou": [
    0.9609375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6469.png",
  "routing_mask_path": "masks/IMG_6469_routing.png",
  "raw_mask_fraction": 0.13482090367011001,
  "routing_mask_fraction": 0.15380658436213993
}

IMG_6470

transient

transient_person · 1 semantic box(es) · raw 16.81% · route 18.82%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6470",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6470",
    "boxes_norm_xyxy": [
      [
        0.7123,
        0.0,
        1.0,
        1.0
      ]
    ],
    "description": "Right- and bottom-clipped pedestrian with backpack and black hand bag.",
    "confidence": 0.99
  },
  "boxes_xyxy": [
    [
      358.99920654296875,
      0.0,
      504.0,
      378.0
    ]
  ],
  "sam_predicted_iou": [
    0.94921875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6470.png",
  "routing_mask_path": "masks/IMG_6470_routing.png",
  "raw_mask_fraction": 0.16813114134542706,
  "routing_mask_fraction": 0.18823486184597296
}

IMG_6471

transient

transient_person · 1 semantic box(es) · raw 10.49% · route 14.30%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6471",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6471",
    "boxes_norm_xyxy": [
      [
        0.0496,
        0.1693,
        0.2877,
        1.0
      ]
    ],
    "description": "Bottom-clipped pedestrian at left, including backpack.",
    "confidence": 0.99
  },
  "boxes_xyxy": [
    [
      24.99839973449707,
      63.995399475097656,
      145.00079345703125,
      378.0
    ]
  ],
  "sam_predicted_iou": [
    0.96484375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6471.png",
  "routing_mask_path": "masks/IMG_6471_routing.png",
  "raw_mask_fraction": 0.10486478542034097,
  "routing_mask_fraction": 0.14297787016040983
}

IMG_6472

transient

transient_person · 1 semantic box(es) · raw 1.97% · route 2.50%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6472",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6472",
    "boxes_norm_xyxy": [
      [
        0.0,
        0.5265,
        0.0873,
        0.8492
      ]
    ],
    "description": "Left-clipped pedestrian fragment dominated by backpack and sleeve.",
    "confidence": 0.9
  },
  "boxes_xyxy": [
    [
      0.0,
      199.01699829101562,
      43.99919891357422,
      320.9975891113281
    ]
  ],
  "sam_predicted_iou": [
    0.97265625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6472.png",
  "routing_mask_path": "masks/IMG_6472_routing.png",
  "raw_mask_fraction": 0.01974153859074494,
  "routing_mask_fraction": 0.0249748047367095
}

IMG_6473

transient

transient_person · 1 semantic box(es) · raw 3.37% · route 4.22%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6473",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6473",
    "boxes_norm_xyxy": [
      [
        0.0,
        0.3836,
        0.1091,
        1.0
      ]
    ],
    "description": "Left- and bottom-clipped pedestrian.",
    "confidence": 0.96
  },
  "boxes_xyxy": [
    [
      0.0,
      145.00079345703125,
      54.98640060424805,
      378.0
    ]
  ],
  "sam_predicted_iou": [
    0.96484375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6473.png",
  "routing_mask_path": "masks/IMG_6473_routing.png",
  "raw_mask_fraction": 0.033651423532375914,
  "routing_mask_fraction": 0.04220206601158982
}

IMG_6474

transient

transient_person · 1 semantic box(es) · raw 20.11% · route 23.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6474",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6474",
    "boxes_norm_xyxy": [
      [
        0.0159,
        0.0,
        0.377,
        1.0
      ]
    ],
    "description": "Large clipped pedestrian with backpack and carried black bag.",
    "confidence": 0.99
  },
  "boxes_xyxy": [
    [
      8.01360034942627,
      0.0,
      190.00799560546875,
      378.0
    ]
  ],
  "sam_predicted_iou": [
    0.94140625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6474.png",
  "routing_mask_path": "masks/IMG_6474_routing.png",
  "raw_mask_fraction": 0.2010686990845721,
  "routing_mask_fraction": 0.23001700680272108
}

IMG_6475

transient

transient_person · 1 semantic box(es) · raw 15.97% · route 20.33%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6475",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6475",
    "boxes_norm_xyxy": [
      [
        0.1508,
        0.0,
        0.6429,
        0.8254
      ]
    ],
    "description": "Top-clipped walking pedestrian, including backpack and large carried bag.",
    "confidence": 0.99
  },
  "boxes_xyxy": [
    [
      76.0031967163086,
      0.0,
      324.0216064453125,
      312.0011901855469
    ]
  ],
  "sam_predicted_iou": [
    0.9453125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6475.png",
  "routing_mask_path": "masks/IMG_6475_routing.png",
  "raw_mask_fraction": 0.15965398505081044,
  "routing_mask_fraction": 0.20334677080708827
}

IMG_6476

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6476",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6476.png",
  "routing_mask_path": "masks/IMG_6476_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6477

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6477",
  "cause": "no_prominent_transient",
  "agent_record": {
    "frame_id": "IMG_6477",
    "boxes_norm_xyxy": [],
    "description": "No transient person; only static fountain, sculpture, buildings, and street fixtures.",
    "confidence": 0.99
  },
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6477.png",
  "routing_mask_path": "masks/IMG_6477_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6478

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6478",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6478.png",
  "routing_mask_path": "masks/IMG_6478_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6479

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6479",
  "cause": "no_prominent_transient",
  "agent_record": {
    "frame_id": "IMG_6479",
    "boxes_norm_xyxy": [],
    "description": "No transient person; close view of static fountain and sculpture.",
    "confidence": 0.99
  },
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6479.png",
  "routing_mask_path": "masks/IMG_6479_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6480

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6480",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6480.png",
  "routing_mask_path": "masks/IMG_6480_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6481

transient

transient_person · 1 semantic box(es) · raw 1.44% · route 2.03%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6481",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6481",
    "boxes_norm_xyxy": [
      [
        0.6687,
        0.0,
        0.869,
        0.1376
      ]
    ],
    "description": "Severely top-clipped pedestrian with backpack/black bag visible above fountain.",
    "confidence": 0.88
  },
  "boxes_xyxy": [
    [
      337.0248107910156,
      0.0,
      437.97601318359375,
      52.01279830932617
    ]
  ],
  "sam_predicted_iou": [
    0.875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6481.png",
  "routing_mask_path": "masks/IMG_6481_routing.png",
  "raw_mask_fraction": 0.014350802049214748,
  "routing_mask_fraction": 0.020271688922482572
}

IMG_6482

transient

transient_person · 1 semantic box(es) · raw 1.87% · route 2.70%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6482",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6482",
    "boxes_norm_xyxy": [
      [
        0.5774,
        0.1878,
        0.7083,
        0.4868
      ]
    ],
    "description": "One background pedestrian, including backpack and black hand bag.",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      291.0096130371094,
      70.9884033203125,
      356.9831848144531,
      184.01040649414062
    ]
  ],
  "sam_predicted_iou": [
    0.9296875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6482.png",
  "routing_mask_path": "masks/IMG_6482_routing.png",
  "raw_mask_fraction": 0.018686486940455195,
  "routing_mask_fraction": 0.027021919879062738
}

IMG_6483

transient

transient_person · 1 semantic box(es) · raw 1.83% · route 2.80%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6483",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6483",
    "boxes_norm_xyxy": [
      [
        0.3651,
        0.4153,
        0.504,
        0.7249
      ]
    ],
    "description": "One pedestrian near road, including backpack and hand bag.",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      184.01040649414062,
      156.9833984375,
      254.01600646972656,
      274.01220703125
    ]
  ],
  "sam_predicted_iou": [
    0.82421875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6483.png",
  "routing_mask_path": "masks/IMG_6483_routing.png",
  "raw_mask_fraction": 0.018277063911984547,
  "routing_mask_fraction": 0.027998236331569663
}

IMG_6484

transient

transient_person · 1 semantic box(es) · raw 2.42% · route 3.67%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6484",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6484",
    "boxes_norm_xyxy": [
      [
        0.119,
        0.4444,
        0.2877,
        0.8016
      ]
    ],
    "description": "One pedestrian at left, including backpack and black hand bag.",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      59.97600173950195,
      167.9832000732422,
      145.00079345703125,
      303.0047912597656
    ]
  ],
  "sam_predicted_iou": [
    0.93359375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6484.png",
  "routing_mask_path": "masks/IMG_6484_routing.png",
  "raw_mask_fraction": 0.0242294448643655,
  "routing_mask_fraction": 0.03672734525909129
}

IMG_6485

transient

transient_person · 1 semantic box(es) · raw 1.75% · route 2.58%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6485",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6485",
    "boxes_norm_xyxy": [
      [
        0.2401,
        0.2302,
        0.3353,
        0.5265
      ]
    ],
    "description": "One background pedestrian, including backpack and black hand bag.",
    "confidence": 0.97
  },
  "boxes_xyxy": [
    [
      121.0103988647461,
      87.0156021118164,
      168.99119567871094,
      199.01699829101562
    ]
  ],
  "sam_predicted_iou": [
    0.91015625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6485.png",
  "routing_mask_path": "masks/IMG_6485_routing.png",
  "raw_mask_fraction": 0.017452968841857732,
  "routing_mask_fraction": 0.025798899806836313
}

IMG_6486

transient

transient_person · 1 semantic box(es) · raw 1.45% · route 2.24%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6486",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6486",
    "boxes_norm_xyxy": [
      [
        0.4365,
        0.0397,
        0.5397,
        0.3148
      ]
    ],
    "description": "Top-near pedestrian, including backpack and black hand bag.",
    "confidence": 0.97
  },
  "boxes_xyxy": [
    [
      219.99600219726562,
      15.006600379943848,
      272.0087890625,
      118.99440002441406
    ]
  ],
  "sam_predicted_iou": [
    0.921875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6486.png",
  "routing_mask_path": "masks/IMG_6486_routing.png",
  "raw_mask_fraction": 0.014545015537079029,
  "routing_mask_fraction": 0.02240278827580415
}

IMG_6487

transient

transient_person · 1 semantic box(es) · raw 0.45% · route 0.83%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6487",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6487",
    "boxes_norm_xyxy": [
      [
        0,
        0,
        0.0595,
        0.1349
      ]
    ],
    "description": "左上角有被画面裁切但可明确识别的近景行人双腿。",
    "confidence": 0.9
  },
  "boxes_xyxy": [
    [
      0.0,
      0.0,
      29.988000869750977,
      50.9921989440918
    ]
  ],
  "sam_predicted_iou": [
    0.92578125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6487.png",
  "routing_mask_path": "masks/IMG_6487_routing.png",
  "raw_mask_fraction": 0.004482657260435038,
  "routing_mask_fraction": 0.008267195767195767
}

IMG_6488

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6488",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6488.png",
  "routing_mask_path": "masks/IMG_6488_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6489

transient

transient_person · 3 semantic box(es) · raw 3.77% · route 5.99%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6489",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6489",
    "boxes_norm_xyxy": [
      [
        0.1171,
        0.0,
        0.2877,
        0.3492
      ],
      [
        0.0893,
        0.0317,
        0.131,
        0.1693
      ],
      [
        0.3532,
        0.0,
        0.3968,
        0.127
      ]
    ],
    "description": "One top-clipped foreground pedestrian with bag and two small background pedestrians.",
    "confidence": 0.82
  },
  "boxes_xyxy": [
    [
      59.01839828491211,
      0.0,
      145.00079345703125,
      131.9976043701172
    ],
    [
      45.007198333740234,
      11.982600212097168,
      66.02400207519531,
      63.995399475097656
    ],
    [
      178.01280212402344,
      0.0,
      199.98719787597656,
      48.00600051879883
    ]
  ],
  "sam_predicted_iou": [
    0.92578125,
    0.890625,
    0.5234375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6489.png",
  "routing_mask_path": "masks/IMG_6489_routing.png",
  "raw_mask_fraction": 0.03765642059292853,
  "routing_mask_fraction": 0.059906987486352564
}

IMG_6490

transient

transient_person · 3 semantic box(es) · raw 6.13% · route 8.61%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6490",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6490",
    "boxes_norm_xyxy": [
      [
        0.0278,
        0.3333,
        0.2361,
        0.9444
      ],
      [
        0.3948,
        0.2831,
        0.4385,
        0.3995
      ],
      [
        0.5992,
        0.1614,
        0.6329,
        0.2302
      ]
    ],
    "description": "One prominent foreground pedestrian with bag and two background pedestrians.",
    "confidence": 0.86
  },
  "boxes_xyxy": [
    [
      14.011199951171875,
      125.98739624023438,
      118.99440002441406,
      356.9831848144531
    ],
    [
      198.9792022705078,
      107.01180267333984,
      221.00399780273438,
      151.01100158691406
    ],
    [
      301.9967956542969,
      61.00920104980469,
      318.9815979003906,
      87.0156021118164
    ]
  ],
  "sam_predicted_iou": [
    0.94921875,
    0.8671875,
    0.74609375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6490.png",
  "routing_mask_path": "masks/IMG_6490_routing.png",
  "raw_mask_fraction": 0.061282228940959096,
  "routing_mask_fraction": 0.08613630637440162
}

IMG_6491

transient

transient_person · 4 semantic box(es) · raw 4.80% · route 6.83%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6491",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6491",
    "boxes_norm_xyxy": [
      [
        0.1488,
        0.6032,
        0.2917,
        1.0
      ],
      [
        0.3135,
        0.5608,
        0.3611,
        0.6878
      ],
      [
        0.3948,
        0.5635,
        0.4345,
        0.6614
      ],
      [
        0.4841,
        0.5106,
        0.5337,
        0.6111
      ]
    ],
    "description": "Bottom-clipped foreground pedestrian plus three distinct pedestrians crossing in background.",
    "confidence": 0.86
  },
  "boxes_xyxy": [
    [
      74.99520111083984,
      228.0095977783203,
      147.0167999267578,
      378.0
    ],
    [
      158.00399780273438,
      211.98240661621094,
      181.99440002441406,
      259.9884033203125
    ],
    [
      198.9792022705078,
      213.0030059814453,
      218.98800659179688,
      250.0092010498047
    ],
    [
      243.9864044189453,
      193.00680541992188,
      268.98480224609375,
      230.9958038330078
    ]
  ],
  "sam_predicted_iou": [
    0.9609375,
    0.77734375,
    0.73828125,
    0.8203125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6491.png",
  "routing_mask_path": "masks/IMG_6491_routing.png",
  "raw_mask_fraction": 0.04799172755521962,
  "routing_mask_fraction": 0.06832640463592844
}

IMG_6492

transient

transient_person · 1 semantic box(es) · raw 0.86% · route 1.32%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6492",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6492",
    "boxes_norm_xyxy": [
      [
        0.3234,
        0,
        0.4683,
        0.0661
      ]
    ],
    "description": "上缘、喷泉后方露出一名被严重遮挡的行人及深色随身物。",
    "confidence": 0.9
  },
  "boxes_xyxy": [
    [
      162.9936065673828,
      0.0,
      236.023193359375,
      24.98579978942871
    ]
  ],
  "sam_predicted_iou": [
    0.625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6492.png",
  "routing_mask_path": "masks/IMG_6492_routing.png",
  "raw_mask_fraction": 0.008603132611069118,
  "routing_mask_fraction": 0.013243260267069791
}

IMG_6493

transient

transient_person · 5 semantic box(es) · raw 4.96% · route 8.18%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6493",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6493",
    "boxes_norm_xyxy": [
      [
        0.0734,
        0.1693,
        0.127,
        0.418
      ],
      [
        0.1052,
        0.1455,
        0.1944,
        0.4074
      ],
      [
        0.2123,
        0.1905,
        0.2698,
        0.3519
      ],
      [
        0.2321,
        0.119,
        0.3909,
        0.4418
      ],
      [
        0.4484,
        0.1534,
        0.506,
        0.3307
      ]
    ],
    "description": "Five pedestrians: two at far left, one center-left, prominent bag-carrying subject, and one crossing at center-right.",
    "confidence": 0.91
  },
  "boxes_xyxy": [
    [
      36.99359893798828,
      63.995399475097656,
      64.00800323486328,
      158.00399780273438
    ],
    [
      53.02080154418945,
      54.999000549316406,
      97.97760009765625,
      153.9971923828125
    ],
    [
      106.99919891357422,
      72.00900268554688,
      135.9792022705078,
      133.01820373535156
    ],
    [
      116.97840118408203,
      44.981998443603516,
      197.0135955810547,
      167.00039672851562
    ],
    [
      225.9936065673828,
      57.985198974609375,
      255.0240020751953,
      125.00460052490234
    ]
  ],
  "sam_predicted_iou": [
    0.84375,
    0.9140625,
    0.90625,
    0.91015625,
    0.8828125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6493.png",
  "routing_mask_path": "masks/IMG_6493_routing.png",
  "raw_mask_fraction": 0.04955068447131939,
  "routing_mask_fraction": 0.08179537246997565
}

IMG_6494

transient

transient_person · 2 semantic box(es) · raw 3.21% · route 4.77%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6494",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6494",
    "boxes_norm_xyxy": [
      [
        0.1151,
        0.2804,
        0.2758,
        0.6296
      ],
      [
        0.3611,
        0.3254,
        0.4067,
        0.4841
      ]
    ],
    "description": "Prominent pedestrian with hand bag plus one background pedestrian crossing behind tree.",
    "confidence": 0.95
  },
  "boxes_xyxy": [
    [
      58.010398864746094,
      105.99120330810547,
      139.00320434570312,
      237.98880004882812
    ],
    [
      181.99440002441406,
      123.0011978149414,
      204.976806640625,
      182.98980712890625
    ]
  ],
  "sam_predicted_iou": [
    0.921875,
    0.890625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6494.png",
  "routing_mask_path": "masks/IMG_6494_routing.png",
  "raw_mask_fraction": 0.03208721760309062,
  "routing_mask_fraction": 0.04769778281683044
}

IMG_6495

transient

transient_person · 6 semantic box(es) · raw 3.96% · route 6.84%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6495",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6495",
    "boxes_norm_xyxy": [
      [
        0.0238,
        0.4233,
        0.0853,
        0.6296
      ],
      [
        0.0575,
        0.418,
        0.1012,
        0.619
      ],
      [
        0.1806,
        0.418,
        0.244,
        0.6296
      ],
      [
        0.2401,
        0.4074,
        0.3373,
        0.7354
      ],
      [
        0.3651,
        0.4365,
        0.4167,
        0.6164
      ],
      [
        0.3948,
        0.4259,
        0.4444,
        0.6085
      ]
    ],
    "description": "Six pedestrians across the street area, including prominent subject with backpack/bag and two overlapping crossing pairs.",
    "confidence": 0.88
  },
  "boxes_xyxy": [
    [
      11.995200157165527,
      160.0074005126953,
      42.9911994934082,
      237.98880004882812
    ],
    [
      28.979999542236328,
      158.00399780273438,
      51.004798889160156,
      233.98199462890625
    ],
    [
      91.02239990234375,
      158.00399780273438,
      122.97599792480469,
      237.98880004882812
    ],
    [
      121.0103988647461,
      153.9971923828125,
      169.99920654296875,
      277.981201171875
    ],
    [
      184.01040649414062,
      164.9969940185547,
      210.0167999267578,
      232.99920654296875
    ],
    [
      198.9792022705078,
      160.99020385742188,
      223.97760009765625,
      230.01300048828125
    ]
  ],
  "sam_predicted_iou": [
    0.87890625,
    0.89453125,
    0.84375,
    0.90234375,
    0.75,
    0.82421875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6495.png",
  "routing_mask_path": "masks/IMG_6495_routing.png",
  "raw_mask_fraction": 0.039551314352901656,
  "routing_mask_fraction": 0.06835789871504157
}

IMG_6496

transient

transient_person · 3 semantic box(es) · raw 2.90% · route 5.03%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6496",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6496",
    "boxes_norm_xyxy": [
      [
        0.0,
        0.5265,
        0.0774,
        0.7725
      ],
      [
        0.4107,
        0.5503,
        0.5317,
        0.8492
      ],
      [
        0.5079,
        0.5608,
        0.5595,
        0.709
      ]
    ],
    "description": "Older pedestrian clipped at left, prominent pedestrian with backpack/bag, and one person immediately behind.",
    "confidence": 0.92
  },
  "boxes_xyxy": [
    [
      0.0,
      199.01699829101562,
      39.00960159301758,
      292.0050048828125
    ],
    [
      206.9927978515625,
      208.01339721679688,
      267.976806640625,
      320.9975891113281
    ],
    [
      255.98159790039062,
      211.98240661621094,
      281.9880065917969,
      268.00201416015625
    ]
  ],
  "sam_predicted_iou": [
    0.94140625,
    0.890625,
    0.84375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6496.png",
  "routing_mask_path": "masks/IMG_6496_routing.png",
  "raw_mask_fraction": 0.029032291929117326,
  "routing_mask_fraction": 0.05031704039640548
}

IMG_6497

transient

transient_person · 3 semantic box(es) · raw 1.74% · route 2.82%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6497",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6497",
    "boxes_norm_xyxy": [
      [
        0.5774,
        0.3069,
        0.621,
        0.4497
      ],
      [
        0.6111,
        0.2989,
        0.6627,
        0.4471
      ],
      [
        0.6369,
        0.2937,
        0.748,
        0.5397
      ]
    ],
    "description": "Three overlapping pedestrians near center-right, including foreground subject and carried bag.",
    "confidence": 0.91
  },
  "boxes_xyxy": [
    [
      291.0096130371094,
      116.0082015991211,
      312.9840087890625,
      169.98660278320312
    ],
    [
      307.994384765625,
      112.98419952392578,
      334.00079345703125,
      169.00379943847656
    ],
    [
      320.9975891113281,
      111.01860046386719,
      376.99200439453125,
      204.00660705566406
    ]
  ],
  "sam_predicted_iou": [
    0.8828125,
    0.87890625,
    0.90625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6497.png",
  "routing_mask_path": "masks/IMG_6497_routing.png",
  "raw_mask_fraction": 0.017421474762744603,
  "routing_mask_fraction": 0.028192449819433946
}

IMG_6498

transient

transient_person · 2 semantic box(es) · raw 0.22% · route 0.75%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6498",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6498",
    "boxes_norm_xyxy": [
      [
        0.5139,
        0.4259,
        0.5595,
        0.5291
      ],
      [
        0.8234,
        0.3862,
        0.873,
        0.5344
      ]
    ],
    "description": "道路侧与右侧车辆旁各有一名虽小但边界可辨的行人。",
    "confidence": 0.9
  },
  "boxes_xyxy": [
    [
      259.005615234375,
      160.99020385742188,
      281.9880065917969,
      199.9998016357422
    ],
    [
      414.99359130859375,
      145.9835968017578,
      439.99200439453125,
      202.00320434570312
    ]
  ],
  "sam_predicted_iou": [
    0.71875,
    0.451171875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6498.png",
  "routing_mask_path": "masks/IMG_6498_routing.png",
  "raw_mask_fraction": 0.0021573444192491812,
  "routing_mask_fraction": 0.00753233392122281
}

IMG_6499

transient

transient_person · 2 semantic box(es) · raw 0.68% · route 1.59%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6499",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6499",
    "boxes_norm_xyxy": [
      [
        0.4246,
        0.2328,
        0.5079,
        0.4101
      ],
      [
        0.8036,
        0.1878,
        0.8611,
        0.381
      ]
    ],
    "description": "喷泉后方及右侧车辆旁各有一名清晰可辨的行人。",
    "confidence": 0.9
  },
  "boxes_xyxy": [
    [
      213.99839782714844,
      87.99839782714844,
      255.98159790039062,
      155.01780700683594
    ],
    [
      405.014404296875,
      70.9884033203125,
      433.994384765625,
      144.01800537109375
    ]
  ],
  "sam_predicted_iou": [
    0.82421875,
    0.78125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6499.png",
  "routing_mask_path": "masks/IMG_6499_routing.png",
  "raw_mask_fraction": 0.006760728982951205,
  "routing_mask_fraction": 0.015941253044427648
}

IMG_6500

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6500",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6500.png",
  "routing_mask_path": "masks/IMG_6500_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6501

transient

transient_person · 3 semantic box(es) · raw 2.93% · route 4.72%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6501",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6501",
    "boxes_norm_xyxy": [
      [
        0.004,
        0.4656,
        0.1607,
        0.836
      ],
      [
        0.4127,
        0.4974,
        0.4464,
        0.5899
      ],
      [
        0.4365,
        0.5053,
        0.4821,
        0.5952
      ]
    ],
    "description": "Prominent pedestrian with bag at left and two cyclists in the background.",
    "confidence": 0.89
  },
  "boxes_xyxy": [
    [
      2.0160000324249268,
      175.99679565429688,
      80.9927978515625,
      316.00799560546875
    ],
    [
      208.00079345703125,
      188.01719665527344,
      224.985595703125,
      222.98219299316406
    ],
    [
      219.99600219726562,
      191.00340270996094,
      242.9783935546875,
      224.985595703125
    ]
  ],
  "sam_predicted_iou": [
    0.91796875,
    0.75,
    0.83203125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6501.png",
  "routing_mask_path": "masks/IMG_6501_routing.png",
  "raw_mask_fraction": 0.029263248509280256,
  "routing_mask_fraction": 0.04715713445872176
}

IMG_6502

transient

transient_person · 2 semantic box(es) · raw 2.43% · route 3.58%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6502",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6502",
    "boxes_norm_xyxy": [
      [
        0.0794,
        0.4656,
        0.2302,
        0.7698
      ],
      [
        0.6746,
        0.5079,
        0.7103,
        0.5899
      ]
    ],
    "description": "One pedestrian with bag at left and one background cyclist.",
    "confidence": 0.94
  },
  "boxes_xyxy": [
    [
      40.017601013183594,
      175.99679565429688,
      116.02079772949219,
      290.9844055175781
    ],
    [
      339.9984130859375,
      191.9862060546875,
      357.9912109375,
      222.98219299316406
    ]
  ],
  "sam_predicted_iou": [
    0.90625,
    0.875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6502.png",
  "routing_mask_path": "masks/IMG_6502_routing.png",
  "raw_mask_fraction": 0.02426618795666415,
  "routing_mask_fraction": 0.035808767951625096
}

IMG_6503

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6503",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6503.png",
  "routing_mask_path": "masks/IMG_6503_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6504

transient

transient_person · 1 semantic box(es) · raw 0.47% · route 0.88%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6504",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6504",
    "boxes_norm_xyxy": [
      [
        0.7063,
        0.0,
        0.7817,
        0.1429
      ]
    ],
    "description": "Severely top-clipped pedestrian visible behind sculpture.",
    "confidence": 0.86
  },
  "boxes_xyxy": [
    [
      355.9751892089844,
      0.0,
      393.976806640625,
      54.01620101928711
    ]
  ],
  "sam_predicted_iou": [
    0.86328125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6504.png",
  "routing_mask_path": "masks/IMG_6504_routing.png",
  "raw_mask_fraction": 0.004718862853783489,
  "routing_mask_fraction": 0.008828840178046526
}

IMG_6505

transient

transient_person · 1 semantic box(es) · raw 0.42% · route 0.66%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6505",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6505",
    "boxes_norm_xyxy": [
      [
        0.9544,
        0.127,
        1.0,
        0.2619
      ]
    ],
    "description": "Severely right-clipped pedestrian fragment with black bag.",
    "confidence": 0.84
  },
  "boxes_xyxy": [
    [
      481.0176086425781,
      48.00600051879883,
      504.0,
      98.99819946289062
    ]
  ],
  "sam_predicted_iou": [
    0.9453125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6505.png",
  "routing_mask_path": "masks/IMG_6505_routing.png",
  "raw_mask_fraction": 0.0042412026539010664,
  "routing_mask_fraction": 0.00655076845553036
}

IMG_6506

transient

transient_person · 1 semantic box(es) · raw 2.89% · route 3.92%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6506",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6506",
    "boxes_norm_xyxy": [
      [
        0.881,
        0.0,
        1.0,
        0.5608
      ]
    ],
    "description": "Right-clipped pedestrian, including large black hand bag.",
    "confidence": 0.96
  },
  "boxes_xyxy": [
    [
      444.02398681640625,
      0.0,
      504.0,
      211.98240661621094
    ]
  ],
  "sam_predicted_iou": [
    0.91796875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6506.png",
  "routing_mask_path": "masks/IMG_6506_routing.png",
  "raw_mask_fraction": 0.028937809691777946,
  "routing_mask_fraction": 0.03922587553539934
}

IMG_6507

transient

transient_person · 1 semantic box(es) · raw 2.29% · route 3.66%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6507",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6507",
    "boxes_norm_xyxy": [
      [
        0.6091,
        0.3995,
        0.7361,
        0.8333
      ]
    ],
    "description": "One pedestrian beside the fountain, including backpack and carried black bag.",
    "confidence": 0.97
  },
  "boxes_xyxy": [
    [
      306.98638916015625,
      151.01100158691406,
      370.994384765625,
      314.9873962402344
    ]
  ],
  "sam_predicted_iou": [
    0.90234375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6507.png",
  "routing_mask_path": "masks/IMG_6507_routing.png",
  "raw_mask_fraction": 0.022948685647098347,
  "routing_mask_fraction": 0.036611866969009824
}

IMG_6508

transient

transient_person · 1 semantic box(es) · raw 2.02% · route 3.54%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6508",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6508",
    "boxes_norm_xyxy": [
      [
        0.4167,
        0.5952,
        0.5794,
        1.0
      ]
    ],
    "description": "One pedestrian partially clipped at the bottom, including backpack and carried bag.",
    "confidence": 0.96
  },
  "boxes_xyxy": [
    [
      210.0167999267578,
      224.985595703125,
      292.0176086425781,
      378.0
    ]
  ],
  "sam_predicted_iou": [
    0.703125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6508.png",
  "routing_mask_path": "masks/IMG_6508_routing.png",
  "raw_mask_fraction": 0.020229696816998405,
  "routing_mask_fraction": 0.035441337028638616
}

IMG_6509

transient

transient_person · 1 semantic box(es) · raw 1.19% · route 1.67%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6509",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6509",
    "boxes_norm_xyxy": [
      [
        0.5516,
        0.0,
        0.6964,
        0.127
      ]
    ],
    "description": "Visible upper fragment of the pedestrian at the top edge, including the black bag.",
    "confidence": 0.91
  },
  "boxes_xyxy": [
    [
      278.00640869140625,
      0.0,
      350.985595703125,
      48.00600051879883
    ]
  ],
  "sam_predicted_iou": [
    0.9296875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6509.png",
  "routing_mask_path": "masks/IMG_6509_routing.png",
  "raw_mask_fraction": 0.011910010917947425,
  "routing_mask_fraction": 0.016718106995884774
}

IMG_6510

transient

transient_person · 1 semantic box(es) · raw 2.20% · route 3.21%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6510",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6510",
    "boxes_norm_xyxy": [
      [
        0.5734,
        0.0714,
        0.7103,
        0.3836
      ]
    ],
    "description": "One pedestrian behind the fountain with a carried black bag and backpack.",
    "confidence": 0.97
  },
  "boxes_xyxy": [
    [
      288.99359130859375,
      26.989200592041016,
      357.9912109375,
      145.00079345703125
    ]
  ],
  "sam_predicted_iou": [
    0.9140625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6510.png",
  "routing_mask_path": "masks/IMG_6510_routing.png",
  "raw_mask_fraction": 0.022003863273704544,
  "routing_mask_fraction": 0.03210296464264718
}

IMG_6511

transient

transient_person · 1 semantic box(es) · raw 2.31% · route 3.28%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6511",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6511",
    "boxes_norm_xyxy": [
      [
        0.623,
        0.2698,
        0.7738,
        0.6058
      ]
    ],
    "description": "One pedestrian with backpack and black hand-carried item.",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      313.99200439453125,
      101.9843978881836,
      389.9952087402344,
      228.99240112304688
    ]
  ],
  "sam_predicted_iou": [
    0.9375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6511.png",
  "routing_mask_path": "masks/IMG_6511_routing.png",
  "raw_mask_fraction": 0.023100907029478458,
  "routing_mask_fraction": 0.03283782648862014
}

IMG_6512

transient

transient_person · 1 semantic box(es) · raw 1.79% · route 2.63%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6512",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6512",
    "boxes_norm_xyxy": [
      [
        0.5437,
        0.4206,
        0.6567,
        0.7302
      ]
    ],
    "description": "One pedestrian walking past the fountain, including backpack.",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      274.0248107910156,
      158.98680114746094,
      330.976806640625,
      276.0155944824219
    ]
  ],
  "sam_predicted_iou": [
    0.91796875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6512.png",
  "routing_mask_path": "masks/IMG_6512_routing.png",
  "raw_mask_fraction": 0.017899134962627027,
  "routing_mask_fraction": 0.02634479717813051
}

IMG_6513

transient

transient_person · 1 semantic box(es) · raw 2.09% · route 3.22%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6513",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6513",
    "boxes_norm_xyxy": [
      [
        0.373,
        0.5344,
        0.5099,
        0.8651
      ]
    ],
    "description": "One pedestrian with backpack and black carried bag.",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      187.99200439453125,
      202.00320434570312,
      256.9895935058594,
      327.0078125
    ]
  ],
  "sam_predicted_iou": [
    0.91015625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6513.png",
  "routing_mask_path": "masks/IMG_6513_routing.png",
  "raw_mask_fraction": 0.020948811623414797,
  "routing_mask_fraction": 0.03218694885361552
}

IMG_6514

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6514",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6514.png",
  "routing_mask_path": "masks/IMG_6514_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6515

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6515",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6515.png",
  "routing_mask_path": "masks/IMG_6515_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6516

transient

transient_person · 1 semantic box(es) · raw 1.27% · route 1.98%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6516",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6516",
    "boxes_norm_xyxy": [
      [
        0.5734,
        0.1746,
        0.6548,
        0.4312
      ]
    ],
    "description": "One pedestrian partly occluded by the fountain structure, including backpack.",
    "confidence": 0.96
  },
  "boxes_xyxy": [
    [
      288.99359130859375,
      65.9988021850586,
      330.0191955566406,
      162.9936065673828
    ]
  ],
  "sam_predicted_iou": [
    0.91015625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6516.png",
  "routing_mask_path": "masks/IMG_6516_routing.png",
  "raw_mask_fraction": 0.012655370790291425,
  "routing_mask_fraction": 0.01979402872260015
}

IMG_6517

transient

transient_person · 1 semantic box(es) · raw 1.08% · route 1.87%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6517",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6517",
    "boxes_norm_xyxy": [
      [
        0.3532,
        0.3439,
        0.4484,
        0.5952
      ]
    ],
    "description": "One pedestrian behind the fountain, including the black carried bag.",
    "confidence": 0.96
  },
  "boxes_xyxy": [
    [
      178.01280212402344,
      129.99420166015625,
      225.9936065673828,
      224.985595703125
    ]
  ],
  "sam_predicted_iou": [
    0.921875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6517.png",
  "routing_mask_path": "masks/IMG_6517_routing.png",
  "raw_mask_fraction": 0.010823465188544554,
  "routing_mask_fraction": 0.018738977072310405
}

IMG_6518

transient

transient_person · 1 semantic box(es) · raw 1.25% · route 2.12%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6518",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6518",
    "boxes_norm_xyxy": [
      [
        0.3294,
        0.2513,
        0.4345,
        0.537
      ]
    ],
    "description": "One pedestrian partly occluded by the arcing pipe, including bag and backpack.",
    "confidence": 0.95
  },
  "boxes_xyxy": [
    [
      166.01759338378906,
      94.99140167236328,
      218.98800659179688,
      202.98599243164062
    ]
  ],
  "sam_predicted_iou": [
    0.84375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6518.png",
  "routing_mask_path": "masks/IMG_6518_routing.png",
  "raw_mask_fraction": 0.01248740236835475,
  "routing_mask_fraction": 0.021164021164021163
}

IMG_6519

transient

transient_person · 1 semantic box(es) · raw 1.91% · route 3.38%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6519",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6519",
    "boxes_norm_xyxy": [
      [
        0.4702,
        0.2831,
        0.6012,
        0.6693
      ]
    ],
    "description": "One pedestrian crossed by the fountain pipe, including black carried bag.",
    "confidence": 0.97
  },
  "boxes_xyxy": [
    [
      236.98080444335938,
      107.01180267333984,
      303.0047912597656,
      252.9954071044922
    ]
  ],
  "sam_predicted_iou": [
    0.8828125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6519.png",
  "routing_mask_path": "masks/IMG_6519_routing.png",
  "raw_mask_fraction": 0.019069664902998235,
  "routing_mask_fraction": 0.033787897875199466
}

IMG_6520

transient

transient_person · 1 semantic box(es) · raw 3.40% · route 5.16%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6520",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6520",
    "boxes_norm_xyxy": [
      [
        0.7282,
        0.0,
        0.9306,
        0.4841
      ]
    ],
    "description": "One pedestrian clipped at the top and right-side area, including backpack and hand bag.",
    "confidence": 0.96
  },
  "boxes_xyxy": [
    [
      367.0127868652344,
      0.0,
      469.02239990234375,
      182.98980712890625
    ]
  ],
  "sam_predicted_iou": [
    0.88671875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6520.png",
  "routing_mask_path": "masks/IMG_6520_routing.png",
  "raw_mask_fraction": 0.0339716133366927,
  "routing_mask_fraction": 0.05156105652137398
}

IMG_6521

transient

transient_person · 1 semantic box(es) · raw 0.30% · route 0.61%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6521",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6521",
    "boxes_norm_xyxy": [
      [
        0.5575,
        0,
        0.6052,
        0.127
      ]
    ],
    "description": "上缘喷泉后方有清晰可辨的行人双腿。",
    "confidence": 0.9
  },
  "boxes_xyxy": [
    [
      280.9800109863281,
      0.0,
      305.02081298828125,
      48.00600051879883
    ]
  ],
  "sam_predicted_iou": [
    0.9140625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6521.png",
  "routing_mask_path": "masks/IMG_6521_routing.png",
  "raw_mask_fraction": 0.0029814394893759974,
  "routing_mask_fraction": 0.006130847400688671
}

IMG_6522

transient

transient_person · 1 semantic box(es) · raw 1.11% · route 1.79%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6522",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6522",
    "boxes_norm_xyxy": [
      [
        0.3492,
        0.2487,
        0.4345,
        0.4815
      ]
    ],
    "description": "One pedestrian substantially occluded by the tree, including visible backpack.",
    "confidence": 0.91
  },
  "boxes_xyxy": [
    [
      175.99679565429688,
      94.00859832763672,
      218.98800659179688,
      182.0070037841797
    ]
  ],
  "sam_predicted_iou": [
    0.87890625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6522.png",
  "routing_mask_path": "masks/IMG_6522_routing.png",
  "raw_mask_fraction": 0.011085915847820609,
  "routing_mask_fraction": 0.017935878054925674
}

IMG_6523

transient

transient_person · 1 semantic box(es) · raw 1.36% · route 2.05%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6523",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6523",
    "boxes_norm_xyxy": [
      [
        0.246,
        0.4418,
        0.3413,
        0.7037
      ]
    ],
    "description": "One pedestrian beside the tree with backpack.",
    "confidence": 0.97
  },
  "boxes_xyxy": [
    [
      123.98400115966797,
      167.00039672851562,
      172.01519775390625,
      265.99859619140625
    ]
  ],
  "sam_predicted_iou": [
    0.93359375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6523.png",
  "routing_mask_path": "masks/IMG_6523_routing.png",
  "raw_mask_fraction": 0.013631687242798354,
  "routing_mask_fraction": 0.0205026455026455
}

IMG_6524

transient

transient_person · 1 semantic box(es) · raw 1.15% · route 1.99%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6524",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6524",
    "boxes_norm_xyxy": [
      [
        0.4504,
        0.4392,
        0.5516,
        0.6905
      ]
    ],
    "description": "One pedestrian beside the tree, including black bag and backpack.",
    "confidence": 0.96
  },
  "boxes_xyxy": [
    [
      227.00160217285156,
      166.01759338378906,
      278.00640869140625,
      261.0090026855469
    ]
  ],
  "sam_predicted_iou": [
    0.9140625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6524.png",
  "routing_mask_path": "masks/IMG_6524_routing.png",
  "raw_mask_fraction": 0.01146909381036365,
  "routing_mask_fraction": 0.019946250104980263
}

IMG_6525

transient

transient_person · 1 semantic box(es) · raw 2.47% · route 3.79%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6525",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6525",
    "boxes_norm_xyxy": [
      [
        0.6468,
        0.172,
        0.7798,
        0.5688
      ]
    ],
    "description": "One pedestrian holding a black item in front, including backpack.",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      325.9872131347656,
      65.01599884033203,
      393.0191955566406,
      215.0063934326172
    ]
  ],
  "sam_predicted_iou": [
    0.921875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6525.png",
  "routing_mask_path": "masks/IMG_6525_routing.png",
  "raw_mask_fraction": 0.024712354077433444,
  "routing_mask_fraction": 0.03794511631813219
}

IMG_6526

transient

transient_person · 1 semantic box(es) · raw 1.86% · route 2.89%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6526",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6526",
    "boxes_norm_xyxy": [
      [
        0.504,
        0.3598,
        0.6409,
        0.619
      ]
    ],
    "description": "One pedestrian partly hidden by the tree and fountain, including carried item and backpack.",
    "confidence": 0.94
  },
  "boxes_xyxy": [
    [
      254.01600646972656,
      136.00439453125,
      323.01361083984375,
      233.98199462890625
    ]
  ],
  "sam_predicted_iou": [
    0.859375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6526.png",
  "routing_mask_path": "masks/IMG_6526_routing.png",
  "raw_mask_fraction": 0.01861824976904342,
  "routing_mask_fraction": 0.0289010665994793
}

IMG_6527

transient

transient_person · 1 semantic box(es) · raw 1.16% · route 2.25%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6527",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6527",
    "boxes_norm_xyxy": [
      [
        0.5198,
        0.1455,
        0.6448,
        0.4048
      ]
    ],
    "description": "One pedestrian behind the tree-side fountain area, including backpack and black item.",
    "confidence": 0.94
  },
  "boxes_xyxy": [
    [
      261.97918701171875,
      54.999000549316406,
      324.97918701171875,
      153.014404296875
    ]
  ],
  "sam_predicted_iou": [
    0.87890625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6527.png",
  "routing_mask_path": "masks/IMG_6527_routing.png",
  "raw_mask_fraction": 0.011579323087259594,
  "routing_mask_fraction": 0.022528764592256657
}

IMG_6528

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6528",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6528.png",
  "routing_mask_path": "masks/IMG_6528_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6529

transient

transient_person · 1 semantic box(es) · raw 0.76% · route 1.16%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6529",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6529",
    "boxes_norm_xyxy": [
      [
        0.629,
        0.0,
        0.7063,
        0.164
      ]
    ],
    "description": "Top-edge fragment of the pedestrian visible behind the fountain.",
    "confidence": 0.89
  },
  "boxes_xyxy": [
    [
      317.0159912109375,
      0.0,
      355.9751892089844,
      61.992000579833984
    ]
  ],
  "sam_predicted_iou": [
    0.94140625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6529.png",
  "routing_mask_path": "masks/IMG_6529_routing.png",
  "raw_mask_fraction": 0.007616318132191148,
  "routing_mask_fraction": 0.011621315192743764
}

IMG_6530

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6530",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6530.png",
  "routing_mask_path": "masks/IMG_6530_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6531

transient

transient_person · 1 semantic box(es) · raw 1.59% · route 2.35%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6531",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6531",
    "boxes_norm_xyxy": [
      [
        0.5655,
        0.0344,
        0.6667,
        0.2989
      ]
    ],
    "description": "One pedestrian behind the fountain, including the raised black carried item.",
    "confidence": 0.96
  },
  "boxes_xyxy": [
    [
      285.0119934082031,
      13.003199577331543,
      336.0168151855469,
      112.98419952392578
    ]
  ],
  "sam_predicted_iou": [
    0.91796875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6531.png",
  "routing_mask_path": "masks/IMG_6531_routing.png",
  "raw_mask_fraction": 0.01586776685983035,
  "routing_mask_fraction": 0.023452590912908373
}

IMG_6532

transient

transient_person · 1 semantic box(es) · raw 1.23% · route 1.98%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6532",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6532",
    "boxes_norm_xyxy": [
      [
        0.5357,
        0.3704,
        0.6389,
        0.6296
      ]
    ],
    "description": "One pedestrian walking away, including backpack and raised black item.",
    "confidence": 0.96
  },
  "boxes_xyxy": [
    [
      269.9927978515625,
      140.01119995117188,
      322.005615234375,
      237.98880004882812
    ]
  ],
  "sam_predicted_iou": [
    0.9296875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6532.png",
  "routing_mask_path": "masks/IMG_6532_routing.png",
  "raw_mask_fraction": 0.01229843789367599,
  "routing_mask_fraction": 0.019773032669858068
}

IMG_6533

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6533",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6533.png",
  "routing_mask_path": "masks/IMG_6533_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6534

transient

transient_person · 1 semantic box(es) · raw 1.42% · route 1.95%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6534",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6534",
    "boxes_norm_xyxy": [
      [
        0.2044,
        0,
        0.369,
        0.1878
      ]
    ],
    "description": "雕塑后方有上半身被画面裁切的近景行人及背包。",
    "confidence": 0.9
  },
  "boxes_xyxy": [
    [
      103.0176010131836,
      0.0,
      185.9759979248047,
      70.9884033203125
    ]
  ],
  "sam_predicted_iou": [
    0.9296875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6534.png",
  "routing_mask_path": "masks/IMG_6534_routing.png",
  "raw_mask_fraction": 0.014182833627278072,
  "routing_mask_fraction": 0.019536827076509618
}

IMG_6535

transient

transient_person · 1 semantic box(es) · raw 2.30% · route 3.86%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6535",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6535",
    "boxes_norm_xyxy": [
      [
        0.1032,
        0.2275,
        0.3075,
        0.5661
      ]
    ],
    "description": "One pedestrian left of the statue, including raised black item and extended arm.",
    "confidence": 0.96
  },
  "boxes_xyxy": [
    [
      52.01279830932617,
      85.99500274658203,
      154.97999572753906,
      213.9857940673828
    ]
  ],
  "sam_predicted_iou": [
    0.90625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6535.png",
  "routing_mask_path": "masks/IMG_6535_routing.png",
  "raw_mask_fraction": 0.02302742084488116,
  "routing_mask_fraction": 0.03862223901906441
}

IMG_6536

transient

transient_person · 1 semantic box(es) · raw 2.75% · route 4.09%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6536",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6536",
    "boxes_norm_xyxy": [
      [
        0.121,
        0.2116,
        0.2579,
        0.664
      ]
    ],
    "description": "One pedestrian partly crossed by the pipe, including backpack and carried black item.",
    "confidence": 0.97
  },
  "boxes_xyxy": [
    [
      60.98400115966797,
      79.98480224609375,
      129.98159790039062,
      250.99200439453125
    ]
  ],
  "sam_predicted_iou": [
    0.921875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6536.png",
  "routing_mask_path": "masks/IMG_6536_routing.png",
  "raw_mask_fraction": 0.027531074158058284,
  "routing_mask_fraction": 0.04093705383387923
}

IMG_6537

transient

transient_person · 1 semantic box(es) · raw 1.12% · route 1.50%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6537",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6537",
    "boxes_norm_xyxy": [
      [
        0.252,
        0,
        0.373,
        0.1534
      ]
    ],
    "description": "雕塑后方有上缘裁切但清晰可辨的行人及背包。",
    "confidence": 0.9
  },
  "boxes_xyxy": [
    [
      127.00800323486328,
      0.0,
      187.99200439453125,
      57.985198974609375
    ]
  ],
  "sam_predicted_iou": [
    0.890625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6537.png",
  "routing_mask_path": "masks/IMG_6537_routing.png",
  "raw_mask_fraction": 0.011217141177458638,
  "routing_mask_fraction": 0.015043671789703535
}

IMG_6538

transient

transient_person · 1 semantic box(es) · raw 1.35% · route 2.25%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6538",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6538",
    "boxes_norm_xyxy": [
      [
        0.2361,
        0.2407,
        0.3492,
        0.5291
      ]
    ],
    "description": "One pedestrian heavily occluded by the static statue, including raised black item.",
    "confidence": 0.91
  },
  "boxes_xyxy": [
    [
      118.99440002441406,
      90.9845962524414,
      175.99679565429688,
      199.9998016357422
    ]
  ],
  "sam_predicted_iou": [
    0.875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6538.png",
  "routing_mask_path": "masks/IMG_6538_routing.png",
  "raw_mask_fraction": 0.013489963886789283,
  "routing_mask_fraction": 0.022492021499958006
}

IMG_6539

transient

transient_person · 1 semantic box(es) · raw 0.30% · route 0.61%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6539",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6539",
    "boxes_norm_xyxy": [
      [
        0.1349,
        0.2434,
        0.1925,
        0.3545
      ]
    ],
    "description": "Small visible head-and-shoulder fragment of the pedestrian behind the statue.",
    "confidence": 0.84
  },
  "boxes_xyxy": [
    [
      67.9896011352539,
      92.00520324707031,
      97.0199966430664,
      134.00100708007812
    ]
  ],
  "sam_predicted_iou": [
    0.89453125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6539.png",
  "routing_mask_path": "masks/IMG_6539_routing.png",
  "raw_mask_fraction": 0.0030234315948601664,
  "routing_mask_fraction": 0.006136096413874192
}

IMG_6540

transient

transient_person · 1 semantic box(es) · raw 1.36% · route 2.19%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6540",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6540",
    "boxes_norm_xyxy": [
      [
        0.0,
        0.2222,
        0.1032,
        0.5053
      ]
    ],
    "description": "One pedestrian clipped by the left image edge and partly hidden by the statue.",
    "confidence": 0.92
  },
  "boxes_xyxy": [
    [
      0.0,
      83.9916000366211,
      52.01279830932617,
      191.00340270996094
    ]
  ],
  "sam_predicted_iou": [
    0.8671875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6540.png",
  "routing_mask_path": "masks/IMG_6540_routing.png",
  "raw_mask_fraction": 0.013558201058201057,
  "routing_mask_fraction": 0.021909381036365163
}

IMG_6541

transient

transient_person · 1 semantic box(es) · raw 1.32% · route 2.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6541",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6541",
    "boxes_norm_xyxy": [
      [
        0.3433,
        0.2619,
        0.4563,
        0.5291
      ]
    ],
    "description": "One pedestrian beside the tree, including backpack and black carried item.",
    "confidence": 0.96
  },
  "boxes_xyxy": [
    [
      173.023193359375,
      98.99819946289062,
      229.97520446777344,
      199.9998016357422
    ]
  ],
  "sam_predicted_iou": [
    0.875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6541.png",
  "routing_mask_path": "masks/IMG_6541_routing.png",
  "raw_mask_fraction": 0.013169774082472495,
  "routing_mask_fraction": 0.020035483329134122
}

IMG_6542

transient

transient_person · 1 semantic box(es) · raw 1.95% · route 3.31%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6542",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6542",
    "boxes_norm_xyxy": [
      [
        0.0813,
        0.1587,
        0.2183,
        0.4762
      ]
    ],
    "description": "One pedestrian partly clipped at left, including backpack and bulky black carried item.",
    "confidence": 0.96
  },
  "boxes_xyxy": [
    [
      40.97520065307617,
      59.98860168457031,
      110.02320098876953,
      180.00360107421875
    ]
  ],
  "sam_predicted_iou": [
    0.90625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6542.png",
  "routing_mask_path": "masks/IMG_6542_routing.png",
  "raw_mask_fraction": 0.019479087931468884,
  "routing_mask_fraction": 0.033131771227009325
}

IMG_6543

transient

transient_person · 1 semantic box(es) · raw 2.02% · route 2.96%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6543",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6543",
    "boxes_norm_xyxy": [
      [
        0.4187,
        0.3228,
        0.5536,
        0.6481
      ]
    ],
    "description": "One pedestrian beside the fountain, including backpack and black hand bag.",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      211.02479553222656,
      122.01840209960938,
      279.014404296875,
      244.98179626464844
    ]
  ],
  "sam_predicted_iou": [
    0.921875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6543.png",
  "routing_mask_path": "masks/IMG_6543_routing.png",
  "raw_mask_fraction": 0.020224447803812884,
  "routing_mask_fraction": 0.02956244226085496
}

IMG_6544

transient

transient_person · 1 semantic box(es) · raw 3.70% · route 5.38%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6544",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6544",
    "boxes_norm_xyxy": [
      [
        0.6012,
        0.2487,
        0.7937,
        0.6746
      ]
    ],
    "description": "One pedestrian crossed by the pipe, including backpack and large black hand bag.",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      303.0047912597656,
      94.00859832763672,
      400.0248107910156,
      254.99879455566406
    ]
  ],
  "sam_predicted_iou": [
    0.921875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6544.png",
  "routing_mask_path": "masks/IMG_6544_routing.png",
  "raw_mask_fraction": 0.03702653901066599,
  "routing_mask_fraction": 0.053786638112034935
}

IMG_6545

transient

transient_person · 1 semantic box(es) · raw 1.16% · route 1.79%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6545",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6545",
    "boxes_norm_xyxy": [
      [
        0.6071,
        0.3942,
        0.7123,
        0.6111
      ]
    ],
    "description": "One pedestrian behind the static statue and fountain, including carried bag.",
    "confidence": 0.94
  },
  "boxes_xyxy": [
    [
      305.9783935546875,
      149.00759887695312,
      358.99920654296875,
      230.9958038330078
    ]
  ],
  "sam_predicted_iou": [
    0.9140625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6545.png",
  "routing_mask_path": "masks/IMG_6545_routing.png",
  "raw_mask_fraction": 0.01164756025867137,
  "routing_mask_fraction": 0.017888636936255985
}

IMG_6546

transient

transient_person · 1 semantic box(es) · raw 0.89% · route 1.53%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6546",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6546",
    "boxes_norm_xyxy": [
      [
        0.4921,
        0.4101,
        0.5873,
        0.627
      ]
    ],
    "description": "One pedestrian partly hidden behind the fountain, including backpack.",
    "confidence": 0.94
  },
  "boxes_xyxy": [
    [
      248.01840209960938,
      155.01780700683594,
      295.99920654296875,
      237.00599670410156
    ]
  ],
  "sam_predicted_iou": [
    0.90625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6546.png",
  "routing_mask_path": "masks/IMG_6546_routing.png",
  "raw_mask_fraction": 0.008855085243974133,
  "routing_mask_fraction": 0.015306122448979591
}

IMG_6547

transient

transient_person · 1 semantic box(es) · raw 1.06% · route 1.58%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6547",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6547",
    "boxes_norm_xyxy": [
      [
        0.3155,
        0.0,
        0.4067,
        0.164
      ]
    ],
    "description": "Top-edge fragment of one pedestrian behind the fountain.",
    "confidence": 0.89
  },
  "boxes_xyxy": [
    [
      159.01199340820312,
      0.0,
      204.976806640625,
      61.992000579833984
    ]
  ],
  "sam_predicted_iou": [
    0.92578125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6547.png",
  "routing_mask_path": "masks/IMG_6547_routing.png",
  "raw_mask_fraction": 0.010639749727051314,
  "routing_mask_fraction": 0.015783782648862012
}

IMG_6548

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6548",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6548.png",
  "routing_mask_path": "masks/IMG_6548_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6549

transient

transient_person · 2 semantic box(es) · raw 1.26% · route 2.28%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6549",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6549",
    "boxes_norm_xyxy": [
      [
        0.1865,
        0.1905,
        0.2837,
        0.4339
      ],
      [
        0.2698,
        0.1958,
        0.2996,
        0.3016
      ]
    ],
    "description": "Main pedestrian with backpack and black bag near the tree, plus one small background pedestrian confirmed transient against the render.",
    "confidence": 0.88
  },
  "boxes_xyxy": [
    [
      93.99600219726562,
      72.00900268554688,
      142.98480224609375,
      164.0142059326172
    ],
    [
      135.9792022705078,
      74.01239776611328,
      150.99839782714844,
      114.00479888916016
    ]
  ],
  "sam_predicted_iou": [
    0.85546875,
    0.87109375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6549.png",
  "routing_mask_path": "masks/IMG_6549_routing.png",
  "raw_mask_fraction": 0.012592382632065172,
  "routing_mask_fraction": 0.022833207357016883
}

IMG_6550

transient

transient_person · 4 semantic box(es) · raw 1.51% · route 2.77%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6550",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6550",
    "boxes_norm_xyxy": [
      [
        0.2698,
        0.545,
        0.3651,
        0.746
      ],
      [
        0.1448,
        0.537,
        0.1687,
        0.6455
      ],
      [
        0.1726,
        0.537,
        0.2024,
        0.6376
      ],
      [
        0.1964,
        0.5317,
        0.2262,
        0.6376
      ]
    ],
    "description": "Main pedestrian with backpack and bags, plus three distinct small background pedestrians removed in the clean render.",
    "confidence": 0.87
  },
  "boxes_xyxy": [
    [
      135.9792022705078,
      206.00999450683594,
      184.01040649414062,
      281.9880065917969
    ],
    [
      72.97920227050781,
      202.98599243164062,
      85.0248031616211,
      243.99899291992188
    ],
    [
      86.99040222167969,
      202.98599243164062,
      102.00959777832031,
      241.01280212402344
    ],
    [
      98.98560333251953,
      200.98260498046875,
      114.00479888916016,
      241.01280212402344
    ]
  ],
  "sam_predicted_iou": [
    0.859375,
    0.83203125,
    0.90625,
    0.90234375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6550.png",
  "routing_mask_path": "masks/IMG_6550_routing.png",
  "raw_mask_fraction": 0.015080414882002183,
  "routing_mask_fraction": 0.02769379356680944
}

IMG_6551

transient

transient_person · 1 semantic box(es) · raw 0.67% · route 1.14%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6551",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6551",
    "boxes_norm_xyxy": [
      [
        0.4921,
        0.3439,
        0.5536,
        0.5238
      ]
    ],
    "description": "One pedestrian behind the fountain holding a black item.",
    "confidence": 0.96
  },
  "boxes_xyxy": [
    [
      248.01840209960938,
      129.99420166015625,
      279.014404296875,
      197.99639892578125
    ]
  ],
  "sam_predicted_iou": [
    0.921875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6551.png",
  "routing_mask_path": "masks/IMG_6551_routing.png",
  "raw_mask_fraction": 0.006729234903838079,
  "routing_mask_fraction": 0.011369362559838751
}

IMG_6552

transient

transient_person · 1 semantic box(es) · raw 0.82% · route 1.39%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6552",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6552",
    "boxes_norm_xyxy": [
      [
        0.369,
        0.0053,
        0.4583,
        0.2037
      ]
    ],
    "description": "喷泉后方有清晰行人,上半身可可靠识别。",
    "confidence": 0.9
  },
  "boxes_xyxy": [
    [
      185.9759979248047,
      2.0034000873565674,
      230.9832000732422,
      76.99859619140625
    ]
  ],
  "sam_predicted_iou": [
    0.89453125
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6552.png",
  "routing_mask_path": "masks/IMG_6552_routing.png",
  "raw_mask_fraction": 0.008214705635340555,
  "routing_mask_fraction": 0.01389413790207441
}

IMG_6553

transient

transient_person · 1 semantic box(es) · raw 0.78% · route 1.31%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6553",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6553",
    "boxes_norm_xyxy": [
      [
        0.4901,
        0.2302,
        0.5774,
        0.3995
      ]
    ],
    "description": "One pedestrian heavily occluded by the fountain, including black carried item.",
    "confidence": 0.93
  },
  "boxes_xyxy": [
    [
      247.01040649414062,
      87.0156021118164,
      291.0096130371094,
      151.01100158691406
    ]
  ],
  "sam_predicted_iou": [
    0.91796875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6553.png",
  "routing_mask_path": "masks/IMG_6553_routing.png",
  "raw_mask_fraction": 0.00781053162005543,
  "routing_mask_fraction": 0.013122532963802806
}

IMG_6554

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6554",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6554.png",
  "routing_mask_path": "masks/IMG_6554_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6555

transient

transient_person · 1 semantic box(es) · raw 0.76% · route 1.23%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6555",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6555",
    "boxes_norm_xyxy": [
      [
        0.2242,
        0.3651,
        0.2956,
        0.5317
      ]
    ],
    "description": "One pedestrian mostly hidden behind the fountain, including black carried item.",
    "confidence": 0.92
  },
  "boxes_xyxy": [
    [
      112.9968032836914,
      138.00779724121094,
      148.98240661621094,
      200.98260498046875
    ]
  ],
  "sam_predicted_iou": [
    0.90625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6555.png",
  "routing_mask_path": "masks/IMG_6555_routing.png",
  "raw_mask_fraction": 0.007569077013521458,
  "routing_mask_fraction": 0.012287939867304946
}

IMG_6556

transient

transient_person · 1 semantic box(es) · raw 0.97% · route 1.59%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6556",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6556",
    "boxes_norm_xyxy": [
      [
        0.3214,
        0.5026,
        0.4107,
        0.6958
      ]
    ],
    "description": "One pedestrian behind the fountain, including the raised black item.",
    "confidence": 0.94
  },
  "boxes_xyxy": [
    [
      161.985595703125,
      189.98280334472656,
      206.9927978515625,
      263.01239013671875
    ]
  ],
  "sam_predicted_iou": [
    0.90625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6556.png",
  "routing_mask_path": "masks/IMG_6556_routing.png",
  "raw_mask_fraction": 0.009673931300915428,
  "routing_mask_fraction": 0.015899260938943477
}

IMG_6557

transient

transient_person · 1 semantic box(es) · raw 0.78% · route 1.30%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6557",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6557",
    "boxes_norm_xyxy": [
      [
        0.6091,
        0.4206,
        0.6885,
        0.6138
      ]
    ],
    "description": "One pedestrian partly hidden by the fountain and statue, including backpack.",
    "confidence": 0.94
  },
  "boxes_xyxy": [
    [
      306.98638916015625,
      158.98680114746094,
      347.0039978027344,
      232.0164031982422
    ]
  ],
  "sam_predicted_iou": [
    0.87109375
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6557.png",
  "routing_mask_path": "masks/IMG_6557_routing.png",
  "raw_mask_fraction": 0.007836776685983034,
  "routing_mask_fraction": 0.012975560594608214
}

IMG_6558

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6558",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6558.png",
  "routing_mask_path": "masks/IMG_6558_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6559

transient

transient_person · 1 semantic box(es) · raw 1.55% · route 2.57%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6559",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6559",
    "boxes_norm_xyxy": [
      [
        0.2698,
        0.3307,
        0.377,
        0.6481
      ]
    ],
    "description": "One fully visible pedestrian with backpack and black carried item.",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      135.9792022705078,
      125.00460052490234,
      190.00799560546875,
      244.98179626464844
    ]
  ],
  "sam_predicted_iou": [
    0.94140625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6559.png",
  "routing_mask_path": "masks/IMG_6559_routing.png",
  "raw_mask_fraction": 0.015468841857730746,
  "routing_mask_fraction": 0.025662425464012765
}

IMG_6560

clean

no_prominent_transient · 0 semantic box(es) · raw 0.00% · route 0.00%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6560",
  "cause": "no_prominent_transient",
  "agent_record": null,
  "boxes_xyxy": [],
  "sam_predicted_iou": [],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6560.png",
  "routing_mask_path": "masks/IMG_6560_routing.png",
  "raw_mask_fraction": 0.0,
  "routing_mask_fraction": 0.0
}

IMG_6561

transient

transient_person · 1 semantic box(es) · raw 24.69% · route 26.06%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6561",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6561",
    "boxes_norm_xyxy": [
      [
        0.0,
        0.0,
        0.3929,
        1.0
      ]
    ],
    "description": "Large close pedestrian clipped by the left, top, and bottom image boundaries, including backpack strap and bulky hand-carried bag.",
    "confidence": 0.98
  },
  "boxes_xyxy": [
    [
      0.0,
      0.0,
      198.0216064453125,
      378.0
    ]
  ],
  "sam_predicted_iou": [
    0.98046875
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6561.png",
  "routing_mask_path": "masks/IMG_6561_routing.png",
  "raw_mask_fraction": 0.2469188292600991,
  "routing_mask_fraction": 0.26057151255563954
}

IMG_6569

transient

transient_person · 1 semantic box(es) · raw 2.14% · route 2.93%

observed target
pose-nearest context
Raw 3DGS train render
Raw residual · auxiliary only
strong-agent semantic boxes
SAM semantic mask
+5 px initial routing mask
boundary candidate +5 px
boundary candidate +9 px
boundary candidate +13 px
boundary candidate +17 px
machine-readable record
{
  "schema_version": "diag.agent_box_sam.v1",
  "frame_id": "IMG_6569",
  "cause": "transient_person",
  "agent_record": {
    "frame_id": "IMG_6569",
    "boxes_norm_xyxy": [
      [
        0.25,
        0.2143,
        0.3631,
        0.5026
      ]
    ],
    "description": "One pedestrian behind the fountain, lower body occluded by the static basin.",
    "confidence": 0.96
  },
  "boxes_xyxy": [
    [
      126.0,
      81.00540161132812,
      183.0023956298828,
      189.98280334472656
    ]
  ],
  "sam_predicted_iou": [
    0.94140625
  ],
  "sam_checkpoint": "/local/scratch-2/fy277/temp/diag3dgs/models/sam2.1/sam2.1_hiera_small.pt",
  "sam_checkpoint_sha256": "6d1aa6f30de5c92224f8172114de081d104bbd23dd9dc5c58996f0cad5dc4d38",
  "base_dilation_px": 5,
  "raw_mask_path": "masks/IMG_6569.png",
  "routing_mask_path": "masks/IMG_6569_routing.png",
  "raw_mask_fraction": 0.021368732678256487,
  "routing_mask_fraction": 0.029341983707063073
}