实验 46:A40 Local Implicit Completion

用 visible-kept surface 与 replacement patch 的 UDF envelope 检验隐式补全是否优于三角面级缝合

实验定位 机制正向 / 非最终高保真

A40 从 A39 的三角面 stitching 转向隐式补全:把 A35 的 visible-kept surface 与 A38 的 TRELLIS.2 raw replacement patch 采样成点云,在局部包围盒中构造 unsigned distance field envelope,再用 marching cubes 导出 mesh。

最佳轻量原型 udf_pitch025_r035 在两个 case 上 watertight rate=100.0%、mean boundary=0、mean nonmanifold=0、mean components=1,相对 A39 snap_bridge 的 component 平均减少 19691.5。代价是 visible p95 drift=0.04150,说明 UDF envelope 会厚化/平滑表面。

实验设计(Image-2 风格)

实验设计图:A40 local implicit completion proxy
实验设计图:A40 local implicit completion proxy
模块设计图:implicit envelope feasibility
模块设计图:implicit envelope feasibility

模块设计(Image-2 风格)

Implicit Proxy

  • 输入:A35 non-conflict kept surface + A38 best replacement patch。
  • 采样:每个 case 约 52k surface points。
  • 距离场:KDTree 最近表面距离,形成 UDF。
  • 重建:在 distance = radius 处 marching cubes。
  • 对照:相对 A39 snap_bridge 看 boundary/component/nonmanifold 是否下降。

Unexpected Issue

Open3D Poisson depth8 原型超过交互运行时间并被停止;A40 改用轻量 UDF marching-cubes。这个选择让实验可复现,但也意味着它是 topology feasibility proxy,不是最终 SDF 高保真补全。

实验结果(表格)

variant 聚合

variant水密率boundarycomponentsnonmanifoldfacesproxy Chamfervisible p95patch p95boundary Δ vs A39component Δ vs A39nonmanifold Δ vs A39
UDF p=.025 r=.035100.0%010183160.035240.041500.04334-10123-19691.5-14240
UDF p=.020 r=.040100.0%01.50290750.040220.046470.04947-10123-19691-14240
UDF p=.025 r=.0500.0%1572.50183440.050180.057340.06096-9966-19690-14240

逐 case 输出

casevariant选择watertightboundarycomponentsnonmanifoldfacesproxy Chamfervisible p95patch p95pitchradiusOBJ
gso_002_input4UDF p=.025 r=.035bestTrue010294200.036020.042230.046240.0250.035/home/jiachen/TRELLIS/experiments/nonorthogonal_gpt_orthoview/outputs/data_survey/vggt_spatial_adapter/support_guidance_a40_local_implicit_completion/gso_002_input4/mesh_udf_pitch025_r035_trellis2_raw.obj
gso_002_input4UDF p=.025 r=.050candidateFalse17630286160.051040.058000.065700.0250.050/home/jiachen/TRELLIS/experiments/nonorthogonal_gpt_orthoview/outputs/data_survey/vggt_spatial_adapter/support_guidance_a40_local_implicit_completion/gso_002_input4/mesh_udf_pitch025_r050_trellis2_raw.obj
gso_002_input4UDF p=.020 r=.040candidateTrue010461300.041030.047050.053140.0200.040/home/jiachen/TRELLIS/experiments/nonorthogonal_gpt_orthoview/outputs/data_survey/vggt_spatial_adapter/support_guidance_a40_local_implicit_completion/gso_002_input4/mesh_udf_pitch020_r040_trellis2_raw.obj
gso_008_input4UDF p=.025 r=.035bestTrue01072120.034450.040760.040440.0250.035/home/jiachen/TRELLIS/experiments/nonorthogonal_gpt_orthoview/outputs/data_survey/vggt_spatial_adapter/support_guidance_a40_local_implicit_completion/gso_008_input4/mesh_udf_pitch025_r035_trellis2_raw.obj
gso_008_input4UDF p=.025 r=.050candidateFalse1382080720.049320.056670.056230.0250.050/home/jiachen/TRELLIS/experiments/nonorthogonal_gpt_orthoview/outputs/data_survey/vggt_spatial_adapter/support_guidance_a40_local_implicit_completion/gso_008_input4/mesh_udf_pitch025_r050_trellis2_raw.obj
gso_008_input4UDF p=.020 r=.040candidateTrue020120200.039400.045890.045810.0200.040/home/jiachen/TRELLIS/experiments/nonorthogonal_gpt_orthoview/outputs/data_survey/vggt_spatial_adapter/support_guidance_a40_local_implicit_completion/gso_008_input4/mesh_udf_pitch020_r040_trellis2_raw.obj

可视化结果

A40 boundary edges
A40 boundary edges
A40 visible drift
A40 visible drift
A40 face budget
A40 face budget
gso_002_input4 implicit completion projection
gso_002_input4 implicit completion projection
gso_008_input4 implicit completion projection
gso_008_input4 implicit completion projection

实验结论

下一步想法