Research map / Point clouds / Point cloud classification
Reliable-View 2D-3D Key-Part Aligned Transformer with Reinforced Masking for 3D Point Cloud Understanding
Xianglong Jin, Zheng Wang, Rong Wang, Feiping Nie
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 multi-view 2D input with KAT (ablation) | Point-MAE baseline · ScanObjectNN OBJ-BG | accuracy (%): 92.94 -> 93.46 | Improved |
| add Reliable Multi-View Selector (RVS) (ablation) | KAT with 3 views · ScanObjectNN OBJ-BG | accuracy (%): 93.46 -> 93.98 | Improved |
| add Reinforced Masking (RM) (ablation) | KAT+RVS · ScanObjectNN OBJ-BG | accuracy (%): 93.98 -> 94.45 | Improved |
| select 3 views instead of 1 (ablation) | KR-MAE · ScanObjectNN OBJ-BG | accuracy (%): 93.98 -> 94.45 | Improved |
| select 6 views instead of 3 (ablation) | KR-MAE · ScanObjectNN OBJ-BG | accuracy (%): 94.45 -> 94.15 | Hurt |
| select 9 views instead of 3 (ablation) | KR-MAE · ScanObjectNN OBJ-BG | accuracy (%): 94.45 -> 93.98 | Hurt |
| use aligned attention instead of MLP fusion (ablation) | KR-MAE · ScanObjectNN OBJ-BG | accuracy (%): 92.60 -> 94.45 | Improved |
| use aligned attention instead of self-attention (ablation) | KR-MAE · ScanObjectNN OBJ-BG | accuracy (%): 93.12 -> 94.45 | Improved |
| use aligned attention instead of standard cross-attention (ablation) | KR-MAE · ScanObjectNN OBJ-BG | accuracy (%): 94.15 -> 94.45 | Improved |
| use reinforced masking instead of random masking, ratio 0.5 (ablation) | Point cloud MAE pre-training · ShapeNet pre-training loss | loss: 0.869 -> 0.790 | Improved |
| use reinforced masking instead of random masking, ratio 0.9 (ablation) | Point cloud MAE pre-training · ShapeNet pre-training loss | loss: 0.889 -> 0.866 | Improved |
| KR-MAE vs Point-MAE on few-shot 10-way 20-shot | KR-MAE · ModelNet40 10-way 20-shot | accuracy (%): 95.0 -> 95.2 | Improved |
Closest papers by meaning
- Point Cloud Self-supervised Learning via 3D to Multi-view Masked Learner (ICCV 2025 · similarity 0.91)
- 3D Feature Prediction for Masked-AutoEncoder-Based Point Cloud Pretraining (ICLR 2024 · similarity 0.89)
- Towards Compact 3D Representations via Point Feature Enhancement Masked Autoencoders (AAAI 2024 · similarity 0.88)
- RI-MAE: Rotation-Invariant Masked AutoEncoders for Self-Supervised Point Cloud Representation Learning (AAAI 2025 · similarity 0.86)
- MM-Point: Multi-View Information-Enhanced Multi-Modal Self-Supervised 3D Point Cloud Understanding (AAAI 2024 · similarity 0.86)
- Multi-Scale Neighborhood Occupancy Masked Autoencoder for Self-Supervised Learning in LiDAR Point Clouds (CVPR 2025 · similarity 0.86)
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.
