Block 01 · C1 · +64 px
Confidence 0.95. The +64 contour already leaves a broad margin around the coat, mannequin neck, and umbrella canopy; all coherent residual and edge halo remain inside.
DiagSplat · local multiradius boundary critic
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。
Confidence 0.95. The +64 contour already leaves a broad margin around the coat, mannequin neck, and umbrella canopy; all coherent residual and edge halo remain inside.
Confidence 0.96. The complete coat and umbrella residual, including the hanging sleeve tip, are comfortably enclosed by C1 with only low-value background outside.
Confidence 0.97. C1 fully contains the narrow support pole, the color-card protrusion, and the bright residual around the oval base.
Confidence 0.98. The stand base and its cyan defocus halo end well before the +64 contour, so a larger candidate is unnecessary.
Hard mask fraction 7.427% → 13.269%;新增 5.842%。 下载最终 mask PNG
Judge ONE frame-0023 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 local transient object indicated by the current mask. Include thin tips, small protrusions, translucent or defocused boundaries, motion/defocus halos, and color/blur trails. The existing mask may under-cover the object. Do not expand merely to include unrelated static background. 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.