Research map / Point clouds / Shape and surface completion
Multi-Angle Point Cloud-VAE: Unsupervised Feature Learning for 3D Point Clouds From Multiple Angles by Joint Self-Reconstruction and Half-to-Half Prediction
Zhizhong Han, Xiyang Wang, Yu-Shen Liu, Matthias Zwicker
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 |
|---|---|---|---|
| add self-reconstruction branch R (ablation) | MAP-VAE · ModelNet10 (ablation, linear SVM) | accuracy (%): 91.63 -> 94.82 | Improved |
| add half-to-half prediction branch P (ablation) | MAP-VAE · ModelNet10 (ablation, linear SVM) | accuracy (%): 92.40 -> 94.82 | Improved |
| add KL variational constraint (ablation) | MAP-VAE · ModelNet10 (ablation, linear SVM) | accuracy (%): 93.17 -> 94.82 | Improved |
| add half-to-half prediction to global-encoder VAE (ablation) | VAE · ModelNet10 (ablation, linear SVM) | accuracy (%): 93.28 -> 94.82 | Improved |
| add half-to-half prediction to global-encoder AE (ablation) | AE · ModelNet10 (ablation, linear SVM) | accuracy (%): 92.29 -> 94.82 | Improved |
| use geodesic instead of Euclidean distance for splitting (ablation) | MAP-VAE · ModelNet10 (ablation, linear SVM) | accuracy (%): 93.61 -> 94.82 | Improved |
| increase beta from 10 to 1000 (ablation) | MAP-VAE · ModelNet10 (parameter sweep) | accuracy (%): 93.72 -> 94.82 | Improved |
| increase beta from 1000 to 10000 (ablation) | MAP-VAE · ModelNet10 (parameter sweep) | accuracy (%): 94.82 -> 93.72 | Hurt |
| use W=6 angles instead of 12 (ablation) | MAP-VAE · ModelNet10 (parameter sweep) | accuracy (%): 93.17 -> 94.82 | Improved |
| use W=6 angles instead of 1 (ablation) | MAP-VAE · ModelNet10 (parameter sweep) | accuracy (%): 92.95 -> 94.82 | Improved |
| place 6 angles uniformly instead of continuously (S-6) (ablation) | MAP-VAE · ModelNet10 (parameter sweep) | accuracy (%): 93.39 -> 94.82 | Improved |
| use alpha 0.01 instead of 0.1 (ablation) | MAP-VAE · ModelNet10 (parameter sweep) | accuracy (%): 92.62 -> 94.82 | Improved |
Closest papers by meaning
- GeoMAE: Masked Geometric Target Prediction for Self-Supervised Point Cloud Pre-Training (CVPR 2023 · similarity 0.90)
- Multi-Scale Part-Based Feature Representation for 3D Domain Generalization and Adaptation (TPAMI 2024 · similarity 0.89)
- Voint Cloud: Multi-View Point Cloud Representation for 3D Understanding (ICLR 2023 · similarity 0.88)
- Shape Self-Correction for Unsupervised Point Cloud Understanding (ICCV 2021 · similarity 0.88)
- PointSea: Point Cloud Completion via Self-structure Augmentation (IJCV 2025 · similarity 0.88)
- Sonata: Self-Supervised Learning of Reliable Point Representations (CVPR 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.
