Research map / Point clouds / Shape and surface completion
LAKe-Net: Topology-Aware Point Cloud Completion by Localizing Aligned Keypoints
Junshu Tang, Zhijun Gong, Ran Yi, Yuan Xie, Lizhuang Ma
Check your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 aligned keypoints and surface-skeleton assistance instead of direct point completion | LAKe-Net · PCN test | CD-L2 (×10^4), average: Snowflake 2.336; LAKe-Net 1.944 | Improved |
| use aligned keypoints and surface-skeleton assistance instead of direct point completion | LAKe-Net · PCN test | EMD (×10^2), average: PointTr 1.764; LAKe-Net 1.557 | Improved |
| use aligned keypoints and surface-skeleton assistance instead of direct point completion | LAKe-Net · ShapeNet55 test | CD-L2 (×10^3), average: GRNet 1.15; LAKe-Net 0.89 | Improved |
| use aligned keypoints and surface-skeleton assistance instead of direct point completion | LAKe-Net · ShapeNet55 test | EMD (×10^3), average: GRNet 28.2; LAKe-Net 31.0 | Hurt |
| add category-specific offset predictors to the keypoint detector (ablation) | LAKe-Net UMKD · KeypointNet, airplane/car/chair/table/vessel | mIoU: Ours-32: 73.7, 60.2, 70.5, 63.2, 62.5; Ours-32 w/o csop: 35.4, 13.2, 27.8, 23.9, 11.0 | Improved |
| increase keypoint count from 16 to 32 (ablation) | LAKe-Net UMKD · KeypointNet, airplane/car/chair/table/vessel | mIoU: 73.2, 58.1, 69.2, 62.5, 61.3 -> 73.7, 60.2, 70.5, 63.2, 62.5 | Improved |
| increase keypoint count from 32 to 64 (ablation) | LAKe-Net UMKD · KeypointNet, airplane/car/chair/table/vessel | mIoU: 73.7, 60.2, 70.5, 63.2, 62.5 -> 74.0, 62.9, 71.3, 65.4, 64.0 | Improved |
| add aligned keypoints instead of using FPS down-sampled ground-truth points (ablation) | LAKe-Net completion · PCN test | EMD (×10^2), average: 1.963 -> 1.557 | Improved |
| add surface-skeleton assistance (ablation) | LAKe-Net completion · PCN test | EMD (×10^2), average: 2.251 -> 1.557 | Improved |
| disturb 5% of detected ground-truth keypoints (ablation) | LAKe-Net completion · PCN test | EMD (×10^2), average: 1.557 -> 1.700 | Hurt |
| misclassify geometrically similar categories (ablation) | LAKe-Net completion · PCN test | EMD (×10^2), average: 1.557 -> 1.590 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- GRNet — GRNet: Gridding Residual Network for Dense Point Cloud Completion (ECCV 2020) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Revisiting Point Cloud Completion: Are We Ready For The Real-World? (ICCV 2025 · similarity 0.91)
- Position-Aware Guided Point Cloud Completion with CLIP Model (AAAI 2025 · similarity 0.90)
- Skeleton-bridged Point Completion: From Global Inference to Local Adjustment (NeurIPS 2020 · similarity 0.90)
- Point Cloud Completion by Skip-Attention Network With Hierarchical Folding (CVPR 2020 · similarity 0.89)
- P2C: Self-Supervised Point Cloud Completion from Single Partial Clouds (ICCV 2023 · similarity 0.89)
- SymmCompletion: High-Fidelity and High-Consistency Point Cloud Completion with Symmetry Guidance (AAAI 2025 · similarity 0.88)
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.
