AI Pulse
📄 论文解读

渲染看着对,3D其实缺了一块

你看到的「分割成功」可能只是渲染出来的假象。现有 3D 高斯分割方法大多靠渲染视图来训练和评估,而渲染时的 alpha 混合会让监督信号集中在贡献最大的那些高斯点上,被遮挡的、贡献小的点几乎没被学到。结果就是:渲染出的遮罩很准,但 3D 里选中的物体其实是残缺的。这篇的做法是绕开「按贡献加权」的梯度传递,用局部深度结构把每个像素背后的高斯点分组,直接监督选中的整个深度簇,让被遮挡的点也拿到信号。他们还造了一个物体移除的合成基准 Removal-3D,用预测深度和真实深度的变化差(RDE)来衡量 3D 选得完不完整;在四个渲染类基线上,这个方法都提升了 3D 完整性,同时在 LERF 和 ScanNet 上不降反升。做 3D 场景编辑、想把物体从重建场景里干净移除的人,这是你明天就可能踩到的坑:别只看渲染图验收。

📄 原文摘要(英文)

3D Gaussian Splatting enables object-level operations on Gaussian primitives, yet existing segmentation methods are largely trained and evaluated through rendered views. This mismatch can produce accurate rendered masks from incomplete 3D selections since alpha-blended feature supervision concentrates learning on dominant contributors while leaving occluded or low-contribution Gaussians weakly supervised. We propose VODA, a plug-and-play supervision objective that exploits local depth structure to group per-pixel contributors and directly supervises all Gaussians in selected depth clusters, bypassing contribution-weighted gradient delivery. To evaluate volumetric completeness, we introduce Removal-3D, a synthetic object-removal benchmark, and the Removal Depth Error (RDE), a metric comparing predicted and ground-truth depth changes. Across four rendering-based baselines, VODA consistently improves volumetric segmentation on Removal-3D while maintaining or improving performance on LERF and ScanNet. Further analysis shows that VODA recovers lower-contribution Gaussians, while simple gradient reweighting and spatial propagation fail to reproduce comparable gains.

arXiv 原文

订阅 AI Pulse

每天 08:00 · 12:30 · 18:30 · 23:50 更新