Research map / Point clouds / Registration and matching
OCA: ODE-Driven Cross-Attention for Image-to-Point-Cloud Registration
Pei An, Jiaqi Yang, Yulong Wang, Quan Siwen, Liangliang Nan
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 OCA module | Matr · 7-Scenes | Registration recall: 0.472 -> 0.552 | Improved |
| add OCA instead of just extra training (Ref) (ablation) | Matr · 7-Scenes | Registration recall: 0.501 -> 0.552 | Improved |
| add OCA instead of just extra training (Ref) (ablation) | Flow-I2P · 7-Scenes | Inlier ratio: 0.459 -> 0.530 | Improved |
| add OCA module on TUM fine-tuning | Matr · TUM | Registration recall: 0.472 -> 0.705 | Improved |
| add OCA module on TUM fine-tuning | LDF-I2P · TUM | Registration recall: 0.636 -> 0.764 | Improved |
| add OCA to CA-I2P on ScanNet (ablation) | CA-I2P · ScanNet (fine-tuned) | Registration recall: 0.739 -> 0.743 | Improved |
| insert OCA without fine-tuning (Zero-OCA) in zero-shot | Flow-I2P · ScanNet (zero-shot) | Registration recall: 0.263 -> 0.414 | Improved |
| insert OCA without fine-tuning (Zero-OCA) in zero-shot | Flow-I2P · ScanNet (zero-shot) | Inlier ratio: 0.310 -> 0.210 | Hurt |
| increase iterations T from 1 to 4 (ablation) | Matr+OCA · 7-Scenes | Inlier ratio: 0.571 -> 0.433 | Hurt |
| increase iterations T from 1 to 4 (ablation) | Matr+OCA · 7-Scenes | Registration recall: 0.489 -> 0.575 | Improved |
| increase iterations T to 5 (ablation) | Matr+OCA · 7-Scenes | Registration recall: 0.575 -> 0.539 | Hurt |
| increase time step tau from 0.10 to 0.25 (ablation) | Matr+OCA · 7-Scenes | Registration recall: 0.552 -> 0.494 | Hurt |
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.
- CA-I2P — CA-I2P: Channel-Adaptive Registration Network with Global Optimal Selection (ICCV 2025) · 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
- FreeReg: Image-to-Point Cloud Registration Leveraging Pretrained Diffusion Models and Monocular Depth Estimators (ICLR 2024 · similarity 0.85)
- Dual Focus-Attention Transformer for Robust Point Cloud Registration (CVPR 2025 · similarity 0.85)
- $\text{I}^2\text{AM}$: Interpreting Image-to-Image Latent Diffusion Models via Bi-Attribution Maps (ICLR 2025 · similarity 0.85)
- PosDiffNet: Positional Neural Diffusion for Point Cloud Registration in a Large Field of View with Perturbations (AAAI 2024 · similarity 0.85)
- Zero-shot RGB-D Point Cloud Registration with Pre-trained Large Vision Model (CVPR 2025 · similarity 0.84)
- Partial Point Cloud Registration with Multi-view 2D Image Learning (AAAI 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.
