DiagSplat · local multiradius boundary critic

Frame 0003 · 四个失败位置 × 四条候选线 → 新 mask

每个 block 的 RGB 和 residual 均作为独立 512×512 PNG 输入。 C1 白=+64，C2 洋红=+128，C3 黄=+192，C4 薄荷绿=+256 px。 选择最小的完整覆盖候选后，以 Gaussian 空间权重平滑组合局部半径，避免方形 crop 接缝。

Residual 仍是有偏辅助：轮廓外 residual 可以否决候选， 但低 residual 不能证明覆盖。最终组合使用 RGB-first 的选择。

Block 01 · C1 · +64 px

Confidence 0.96. The white boundary encloses the left sleeve, soft garment edge, and all coherent residual outside the original mask.

Separate critic input A · RGB · four candidate lines
Separate critic input B · biased residual · four candidate lines

Block 02 · C1 · +64 px

Confidence 0.96. The white boundary contains the right sleeve silhouette and its visible blur/residual halo.

Separate critic input A · RGB · four candidate lines
Separate critic input B · biased residual · four candidate lines

Block 03 · C1 · +64 px

Confidence 0.98. The white boundary contains the full lower-left hem and the small protruding corner below the earlier mask.

Separate critic input A · RGB · four candidate lines
Separate critic input B · biased residual · four candidate lines

Block 04 · C1 · +64 px

Confidence 0.98. The white boundary safely encloses the lower-right hem, inter-panel gap, and residual edge.

Separate critic input A · RGB · four candidate lines
Separate critic input B · biased residual · four candidate lines

四个判断组合得到的新 mask

Hard mask fraction 12.972% → 18.016%；新增 5.044%。 下载最终 mask PNG

old-mask
new-mask
added-hard-region
spatially-smoothed-radius-field
new-mask-on-biased-residual
完整 critic prompt
Judge ONE frame-0003 boundary location at a time.

You receive two separate 512x512 PNG images for the same location:
A. true observed RGB;
B. biased Round-1 residual.
They are separate visual inputs, never a merged composite.

Four thin candidate hard-mask boundaries appear in both images:
C1 white/black = +64 px
C2 magenta/black = +128 px
C3 yellow/black = +192 px
C4 mint/black = +256 px

Choose the SMALLEST candidate that encloses every visible part of the very-near
defocused parasol: paper, wooden ribs and tips, translucent fringe, defocus
halo, and color/blur trail.  If even C4 fails, return INSUFFICIENT.

Residual construction: mean absolute RGB difference between dirty observed and
the Round-1 3DGS render, Gaussian sigma 1.8, clipped at 0.30, Turbo colormap.
The render was trained with an earlier undercovered mask and can reproduce
leaked transient pixels. Therefore coherent residual outside a line proves
that line insufficient, but low residual never proves coverage.

Return: selected candidate, radius, confidence, and one concise RGB-first
reason.