Research map / Point clouds / Shape and surface completion
Unsupervised Learning of Category-Specific Symmetric 3D Keypoints from Point Sets
Clara Fernandez-Labrador, Ajad Chhatkuli, Danda Pani Paudel, Jose J. Guerrero, Cédric Demonceaux, Luc Van Gool
Read the paperCheck 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 category-specific symmetric linear basis keypoint model instead of per-category USIP baseline | symmetric linear basis keypoint network · 8 categories | coverage: 29.4% more coverage on average | Improved |
| use symmetric linear basis model instead of USIP baseline | symmetric linear basis keypoint network · 8 categories | inclusivity: ~95% vs ~89% on average | Mixed |
| use ordered keypoints from the non-rigid basis model instead of USIP baseline | symmetric linear basis keypoint network · 8 categories | correspondence: 100 vs 30.11 to 55.6 | Improved |
| add instance-wise or basis symmetry modelling during training (ablation) | symmetric linear basis keypoint network · chair, 10 keypoints, +-45 degree misalignment | relative rotation error | Improved |
| train with Chamfer loss instead of l2 loss between nodes and keypoints (ablation) | symmetric linear basis keypoint network · keypoint learning | convergence | Improved |
| vary the number of shape basis K between 5 and 10 (ablation) | symmetric linear basis keypoint network · keypoint learning | keypoint quality | No clear effect |
| approximate volume with 3D bounding box in coverage loss (ablation) | symmetric linear basis keypoint network · keypoint learning | performance | No clear effect |
| learn a common category rotation tied to the symmetry plane (ablation) | symmetric linear basis keypoint network · misaligned point clouds | rotation ambiguity | Improved |
Closest papers by meaning
- Unsupervised 3D Structure Inference from Category-Specific Image Collections (CVPR 2024 · similarity 0.93)
- KeypointDeformer: Unsupervised 3D Keypoint Discovery for Shape Control (CVPR 2021 · similarity 0.91)
- Unsupervised Learning of 3D Semantic Keypoints with Mutual Reconstruction (ECCV 2022 · similarity 0.91)
- Discovery of Latent 3D Keypoints via End-to-end Geometric Reasoning (NeurIPS 2018 · similarity 0.90)
- Key-Grid: Unsupervised 3D Keypoints Detection using Grid Heatmap Features (NeurIPS 2024 · similarity 0.87)
- KP-RED: Exploiting Semantic Keypoints for Joint 3D Shape Retrieval and Deformation (CVPR 2024 · 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.
