Research map / Point clouds / Registration and matching
RayI2P: Learning Rays for Image-to-Point Cloud Registration
Xinjun Li, Wenfei Yang, Zhixin Cheng, Jiacheng Deng, Fei Wang, Chen Qian, Tianzhu Zhang
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 patch rays to fused patch features (ablation) | RayI2P pose module · KITTI | Acc (%): 94.48 -> 99.41 | Improved |
| add patch rays to fused patch features (ablation) | RayI2P pose module · nuScenes | Acc (%): 89.79 -> 94.25 | Improved |
| add reference rays to fused features and patch rays (ablation) | RayI2P pose module · nuScenes | Acc (%): 94.25 -> 96.61 | Improved |
| add reference rays to fused features and patch rays (ablation) | RayI2P pose module · KITTI | Acc (%): 99.41 -> 99.75 | Improved |
| replace classical pose solver with learned ray-guided regression (ablation) | RayI2P pose module · nuScenes | RRE (deg): 2.51±20.93 -> 1.89±6.32 | Improved |
| replace classical pose solver with learned ray-guided regression (ablation) | RayI2P pose module · KITTI | Acc (%): 99.62 -> 99.64 | Improved |
| add fused patch features to patch and reference rays (ablation) | RayI2P pose module · nuScenes | Acc (%): 95.38 -> 96.61 | Improved |
| add focus loss (sigma 32) (ablation) | RayI2P · KITTI | RRE (deg): 1.02 -> 0.63 | Improved |
| add focus loss (sigma 32) (ablation) | RayI2P · nuScenes | Acc (%): 94.44 -> 96.61 | Improved |
| increase focus radius sigma from 8 to 32 (ablation) | RayI2P · nuScenes | Acc (%): 94.70 -> 96.61 | Improved |
| increase focus radius sigma from 32 to 256 (ablation) | RayI2P · nuScenes | Acc (%): 96.61 -> 95.39 | Hurt |
| use patch size 8 instead of 4 (ablation) | RayI2P · KITTI | RTE (m): 0.12 -> 0.09 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- GraphI2P — GraphI2P: Image-to-Point Cloud Registration with Exploring Pattern of Correspondence via Graph Learning (CVPR 2025) · better on 5 of 6 shared results
- FreeReg — FreeReg: Image-to-Point Cloud Registration Leveraging Pretrained Diffusion Models and Monocular Depth Estimators (ICLR 2024) · 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
- $\pi^2$: A Simple Framework for 2D-to-3D Registration (NeurIPS 2026 · similarity 0.91)
- GraphI2P: Image-to-Point Cloud Registration with Exploring Pattern of Correspondence via Graph Learning (CVPR 2025 · similarity 0.88)
- 3D Focusing-and-Matching Network for Multi-Instance Point Cloud Registration (NeurIPS 2024 · similarity 0.87)
- Implicit Correspondence Learning for Image-to-Point Cloud Registration (CVPR 2025 · similarity 0.87)
- Generative Point Cloud Registration (ICML 2025 · similarity 0.87)
- PointMC: Multi-instance Point Cloud Registration based on Maximal Cliques (ICML 2024 · 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.
