Research map / Label-efficient and robust learning / Domain adaptation and generalization
CounterPC: Counterfactual Feature Realignment for Unsupervised Domain Adaptation on Point Clouds
Feng Yang, Yichao Cao, Xiu Su, Dan Niu, Xuanpeng Li
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-supervised autoregressive generation (SSG) (ablation) | PointGPT-S · PointDA-10 Avg. | accuracy (%): 73.4 -> 73.9 | Improved |
| add counterfactual feature generation (CFG) (ablation) | CounterPC-S · PointDA-10 Avg. | accuracy (%): 73.9 -> 75.5 | Improved |
| add estimation of target domain distribution (ETD) (ablation) | CounterPC-S · PointDA-10 Avg. | accuracy (%): 75.5 -> 77.5 | Improved |
| add CFG to pseudo-label self-training (ablation) | CounterPC-S · PointDA-10 M→S | accuracy (%): 85.72 -> 86.24 | Improved |
| add ETD to pseudo-label self-training (ablation) | CounterPC-S · PointDA-10 M→S | accuracy (%): 85.72 -> 89.04 | Improved |
| add ETD to pseudo-label self-training (ablation) | CounterPC-S · GraspNetPC-10 K→R | accuracy (%): 76.53 -> 83.74 | Improved |
| raise source CE weight from 0.5 to 2 (ablation) | CounterPC-S · PointDA-10 M→S | accuracy (%): 86.54 -> 89.23 | Improved |
| raise source CE weight from 2 to 4 (ablation) | CounterPC-S · PointDA-10 S→M | accuracy (%): 89.87 -> 86.54 | Hurt |
| raise source CD weight from 5 to 10 (ablation) | CounterPC-S · PointDA-10 S→M | accuracy (%): 88.37 -> 85.47 | Hurt |
| raise target CE weight from 1 to 4 (ablation) | CounterPC-S · PointDA-10 S→M | accuracy (%): 89.87 -> 86.72 | Hurt |
| raise target CD weight from 5 to 10 (ablation) | CounterPC-S · PointDA-10 S→M | accuracy (%): 89.87 -> 86.30 | Hurt |
| CounterPC-S vs MLGR on PointDA-10 | CounterPC · PointDA-10 Avg. | accuracy (%): 2.3 and 2.7 | Improved |
Closest papers by meaning
- PointDAN: A Multi-Scale 3D Domain Adaption Network for Point Cloud Representation (NeurIPS 2019 · similarity 0.87)
- Progressive Classifier and Feature Extractor Adaptation for Unsupervised Domain Adaptation on Point Clouds (ECCV 2024 · similarity 0.87)
- Domain Adaptation on Point Clouds via Geometry-Aware Implicits (CVPR 2022 · similarity 0.87)
- PC-Adapter: Topology-Aware Adapter for Efficient Domain Adaption on Point Clouds with Rectified Pseudo-label (ICCV 2023 · similarity 0.86)
- CL3D: Unsupervised Domain Adaptation for Cross-LiDAR 3D Detection (AAAI 2023 · similarity 0.86)
- Dataset Distillation of 3D Point Clouds via Distribution Matching (NeurIPS 2025 · similarity 0.84)
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.
