Research map / Point clouds / Registration and matching
Rotation-Invariant Transformer for Point Cloud Matching
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 RoITr's intrinsically rotation-invariant architecture | point cloud matching · Rotated 3DLoMatch | Inlier Ratio (%): 40.0 -> 53.2 | Helped |
| Use RoITr's intrinsically rotation-invariant architecture | point cloud matching · Rotated 3DLoMatch | Registration Recall (%): 71.8 -> 77.2 | Helped |
| Use RoITr's intrinsically rotation-invariant architecture | point cloud matching · 4DMatch | NFMR (%): 83.7 -> 83.0 | Mixed |
| Use RoITr's intrinsically rotation-invariant architecture | point cloud matching · 4DLoMatch | NFMR (%): 66.9 -> 69.4 | Helped |
| Add PPF local coordinates to PointTransformer (ablation) | local point-cloud matching backbone · Rotated 3DLoMatch | Inlier Ratio (%): 34.7 -> 49.4 | Helped |
| Replace PointTransformer with the full PPFTrans design (ablation) | local point-cloud matching backbone · Rotated 3DLoMatch | Inlier Ratio (%): 49.4 -> 53.2 | Helped |
| Add attentional abstraction rather than max pooling (ablation) | RoITr · Rotated 3DLoMatch | Inlier Ratio (%): 50.2 -> 53.2 | Helped |
| Add attentional abstraction rather than average pooling (ablation) | RoITr · Rotated 3DLoMatch | Inlier Ratio (%): 52.5 -> 53.2 | Helped |
| Use PPFTrans instead of KPConv (ablation) | RoITr backbone · Rotated 3DLoMatch | Inlier Ratio (%): 42.3 -> 53.2 | Helped |
| Add RoITr's cross-frame position-aware global transformer instead of GeoTrans's global transformer (ablation) | RoITr · Rotated 3DLoMatch | Registration Recall (%): 75.1 -> 77.2 | Helped |
| Add global transformers rather than using none (ablation) | RoITr · Rotated 3DLoMatch | Inlier Ratio (%): 37.6 -> 53.2 | Helped |
| Increase the number of global transformers from one to three (ablation) | RoITr · Rotated 3DLoMatch | Inlier Ratio (%): 42.1 -> 53.2 | Helped |
Closest papers by meaning
- DV-Matcher: Deformation-based Non-rigid Point Cloud Matching Guided by Pre-trained Visual Features (CVPR 2025 · similarity 0.89)
- A Consistency-Aware Spot-Guided Transformer for Versatile and Hierarchical Point Cloud Registration (NeurIPS 2024 · similarity 0.88)
- RI-MAE: Rotation-Invariant Masked AutoEncoders for Self-Supervised Point Cloud Representation Learning (AAAI 2025 · similarity 0.88)
- APML: Adaptive Probabilistic Matching Loss for Robust 3D Point Cloud Reconstruction (NeurIPS 2025 · similarity 0.87)
- Local-consistent Transformation Learning for Rotation-invariant Point Cloud Analysis (CVPR 2024 · similarity 0.87)
- Point Transformer V3: Simpler Faster Stronger (CVPR 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.