Research map / Point clouds / Segmentation and detection
Learning Transferable Features for Point Cloud Detection via 3D Contrastive Co-training
Read the paperCheck 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 domain-specific encoders instead of shared encoder (ablation) | 3D-CoCo · nuScenes->KITTI target | AP3D: 59.1 -> 65.6 | Helped |
| share the BEV transformation module across domains instead of separate modules (ablation) | 3D-CoCo · nuScenes->KITTI target | AP3D: 60.3 -> 65.6 | Helped |
| add balanced background sampling (ablation) | 3D-CoCo · nuScenes->KITTI | AP3D: 62.1 -> 63.3 | Helped |
| add similarity-priority positive selection (ablation) | 3D-CoCo · nuScenes->KITTI | AP3D: 63.3 -> 65.6 | Helped |
| use instance-level instead of prototype-level alignment (ablation) | 3D-CoCo · nuScenes->KITTI | AP3D: 59.2 -> 65.6 | Helped |
| add hard sample mining (uniform dropping + occlusion simulation) (ablation) | 3D-CoCo · nuScenes->KITTI | AP3D: 58.8 -> 65.6 | Helped |
| add uniform point dropping (ablation) | 3D-CoCo · nuScenes->KITTI | AP3D: 58.8 -> 61.3 | Helped |
| use 3D-CoCo instead of ST3D | 3D-CoCo · nuScenes->KITTI VoxelNet | AP3D: 38.3 -> 65.6 | Helped |
| apply SN statistical normalization | SN · nuScenes->KITTI VoxelNet | AP_BEV closed gap: -30.13% | Hurt |
| use co-training instead of self-training | 3D-CoCo · nuScenes->KITTI | AP3D / robustness to pseudo-label quality | Helped |
Closest papers by meaning
- CO3: Cooperative Unsupervised 3D Representation Learning for Autonomous Driving (ICLR 2023 · similarity 0.90)
- GBlobs: Explicit Local Structure via Gaussian Blobs for Improved Cross-Domain LiDAR-based 3D Object Detection (CVPR 2025 · similarity 0.88)
- Towards Learning Group-Equivariant Features for Domain Adaptive 3D Detection (NeurIPS 2024 · similarity 0.88)
- An Empirical Study of the Generalization Ability of Lidar 3D Object Detectors to Unseen Domains (CVPR 2024 · similarity 0.87)
- Point-Cache: Test-time Dynamic and Hierarchical Cache for Robust and Generalizable Point Cloud Analysis (CVPR 2025 · similarity 0.87)
- Mining and Transferring Feature-Geometry Coherence for Unsupervised Point Cloud Registration (NeurIPS 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.