Research map / Point clouds / Point cloud classification
Unsupervised Point Cloud Pre-Training via Occlusion Completion
Hanchen Wang, Qi Liu, Xiangyu Yue, Joan Lasenby, Matt J. Kusner
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 OcCo pre-training | PointNet · ModelNet40 few-shot 5-way 10-shot | accuracy: 52.0 -> 89.7 | Improved |
| add OcCo pre-training | DGCNN · ModelNet40 few-shot 5-way 10-shot | accuracy: 31.6 -> 90.6 | Improved |
| add OcCo pre-training | DGCNN · ScanObjectNN few-shot 5-way 10-shot | accuracy: 62.0 -> 72.4 | Improved |
| add OcCo pre-training | DGCNN · ScanObjectNN classification | accuracy: 82.4 -> 83.9 | Improved |
| add OcCo pre-training | PointNet · ScanObjectNN classification | accuracy: 73.5 -> 80.0 | Improved |
| add Jigsaw pre-training | DGCNN · ModelNet classification | accuracy: 92.5 -> 92.3 | No clear effect |
| add OcCo pre-training | DGCNN · ShapeNetPart | mIoU: 84.4 -> 85.0 | Improved |
| add OcCo pre-training | DGCNN · S3DIS | mIoU: 54.9 -> 58.0 | Improved |
| add OcCo pre-training | DGCNN · SensatUrban | mIoU: 51.96 -> 53.31 | Improved |
| use more pre-training views (1 -> 20) (ablation) | DGCNN OcCo · ScanObjectNN few-shot 10-way 10-shot | accuracy: 42.7 -> 57.0 | Improved |
| use more views beyond 5 (5 -> 20) (ablation) | DGCNN OcCo · ScanObjectNN few-shot 10-way 10-shot | accuracy: 56.9 -> 57.0 | No clear effect |
| increase pre-training categories (1 -> 40) (ablation) | DGCNN OcCo · ScanObjectNN few-shot 10-way 10-shot | accuracy: 37.9 -> 56.8 | Improved |
Closest papers by meaning
- Point Cloud Pre-training with Diffusion Models (CVPR 2024 · similarity 0.91)
- PointGLR: Unsupervised Structural Representation Learning of 3D Point Clouds (TPAMI 2022 · similarity 0.90)
- Masked Discrimination for Self-Supervised Learning on Point Clouds (ECCV 2022 · similarity 0.90)
- FreePoint: Unsupervised Point Cloud Instance Segmentation (CVPR 2024 · similarity 0.90)
- Occlusion-aware Text-Image-Point Cloud Pretraining for Open-World 3D Object Recognition (CVPR 2025 · similarity 0.90)
- Multi-View Representation is What You Need for Point-Cloud Pre-Training (ICLR 2024 · similarity 0.89)
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.
