Research map / Point clouds / Shape and surface completion
POCO: Point Convolution for Surface Reconstruction
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use point convolutions with FKAConv instead of Residual PointNet (ablation) | POCO · ShapeNet chairs train, 13-class generalization | IoU: 0.661 -> 0.882 | Helped |
| Increase interpolation neighbors from 1 to 64 (ablation) | POCO · ShapeNet chairs train, 13-class generalization | IoU: 0.799 -> 0.882 | Helped |
| Increase interpolation neighbors from 64 to 128 (ablation) | POCO · ShapeNet chairs train, 13-class generalization | CD: 4.069 -> 3.611 | Helped |
| Use multi-head rather than single-head attention (ablation) | POCO · ShapeNet chairs train, 13-class generalization | IoU: 0.879 -> 0.895 | Helped |
| Interpolate relative rather than global features (ablation) | POCO · ShapeNet chairs train, 13-class generalization | IoU: 0.854 -> 0.882 | Helped |
| Use learned multi-head attention rather than mean interpolation of relative features (ablation) | POCO · ShapeNet chairs train, 13-class generalization | IoU: 0.883 -> 0.895 | Helped |
| Apply latent-level test-time augmentation with 10 views (ablation) | POCO · ABC, Famous, and Thingi10k | CD: 1.87 -> 1.77; 1.56 -> 1.40; 1.47 -> 1.39 | Helped |
| Use POCO instead of DP-ConvONet | ShapeNet · ShapeNet | IoU: 0.895 -> 0.926 | Helped |
| Use POCO instead of ConvONet | Synthetic Rooms · Synthetic Rooms | IoU: 0.849 -> 0.884 | Helped |
| Use POCO instead of LIG | SceneNet reconstruction · SceneNet, 20 pts/m² | CD: 1.52 -> 0.84 | Helped |
| Use POCO instead of LIG | SceneNet reconstruction · SceneNet, 100 pts/m² | CD: 0.97 -> 0.57 | Helped |
Closest papers by meaning
- OffsetOPT: Explicit Surface Reconstruction without Normals (CVPR 2025 · similarity 0.87)
- Arbitrary-Scale Point Cloud Upsampling by Voxel-Based Network with Latent Geometric-Consistent Learning (AAAI 2024 · similarity 0.87)
- SPAR3D: Stable Point-Aware Reconstruction of 3D Objects from Single Images (CVPR 2025 · similarity 0.87)
- GridFormer: Point-Grid Transformer for Surface Reconstruction (AAAI 2024 · similarity 0.87)
- LISR: Learning Linear 3D Implicit Surface Representation Using Compactly Supported Radial Basis Functions (AAAI 2024 · similarity 0.87)
- Matérn Kernels for Tunable Implicit Surface Reconstruction (ICLR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.