Research map / Point clouds / Segmentation and detection
Point Cloud Domain Adaptation via Masked Local 3D Structure Prediction
Hanxue Liang, Hehe Fan, Zhiwen Fan, Yi Wang, Tianlong Chen, Yu Cheng, Zhangyang Wang
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 MLSP self-supervision on target data | DGCNN baseline · PointDA-10 | accuracy (%): 62.2 -> 71.0 | Improved |
| add cardinality prediction alone (ablation) | DGCNN baseline · PointDA-10 | accuracy (%): 62.2 -> 68.7 | Improved |
| add position prediction alone (ablation) | DGCNN baseline · PointDA-10 | accuracy (%): 62.2 -> 68.6 | Improved |
| add normal prediction alone (ablation) | DGCNN baseline · PointDA-10 | accuracy (%): 62.2 -> 68.2 | Improved |
| combine cardinality, position and normal prediction (ablation) | MLSP · PointDA-10 | accuracy (%): 68.7 -> 71.0 | Improved |
| use two-point KL cardinality loss instead of cross-entropy (ablation) | MLSP cardinality only · PointDA-10 | accuracy (%): 67.8 -> 68.7 | Improved |
| increase cardinality classes beyond 8 (ablation) | MLSP · PointDA-10 | accuracy (%) | Hurt |
| use more than 15 neighbours for normal estimation (ablation) | MLSP · PointDA-10 | accuracy (%) | Hurt |
| add entropy-guided self-paced alignment (EGA) | MLSP · PointDA-10 | accuracy (%): 71.0 -> 74.0 | Improved |
| select pseudo labels by entropy (EGA) instead of probability threshold (SPST) | MLSP · PointDA-10 | accuracy (%): 73.8 -> 74.0 | Improved |
| apply MLSP to segmentation | DGCNN · PointSegDA | mean IoU (%): 58.1 -> 59.5 vs DefRec | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PointDAN — PointDAN: A Multi-Scale 3D Domain Adaption Network for Point Cloud Representation (NeurIPS 2019) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Self-Supervised Global-Local Structure Modeling for Point Cloud Domain Adaptation With Reliable Voted Pseudo Labels (CVPR 2022 · similarity 0.90)
- Mining Point Cloud Local Structures by Kernel Correlation and Graph Pooling (CVPR 2018 · similarity 0.90)
- Masked Clustering Prediction for Unsupervised Point Cloud Pre-training (AAAI 2026 · similarity 0.90)
- A Linear Time and Space Local Point Cloud Geometry Encoder via Vectorized Kernel Mixture (VecKM) (ICML 2024 · similarity 0.88)
- X-3D: Explicit 3D Structure Modeling for Point Cloud Recognition (CVPR 2024 · similarity 0.88)
- Taylor Series-Inspired Local Structure Fitting Network for Few-shot Point Cloud Semantic Segmentation (AAAI 2025 · similarity 0.88)
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.
