Research map / Point clouds / Registration and matching
DeepI2P: Image-to-Point Cloud Registration via Deep Classification
Check 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 frustum classification with inverse projection instead of direct pose regression | DeepI2P · Oxford | RTE (m): 5.02\pm2.89 -> 1.65\pm1.36 | Helped |
| Use frustum classification with inverse projection instead of direct pose regression | DeepI2P · KITTI | RTE (m): 4.94\pm2.87 -> 3.28\pm3.09 | Helped |
| Constrain inverse projection to 3-DoF rather than optimize full 6-DoF pose (ablation) | Frus. Cls. + Inv.Proj. · Oxford | RRE (°): 15.00\pm13.64 -> 4.14\pm4.90 | Helped |
| Constrain inverse projection to 3-DoF rather than optimize full 6-DoF pose (ablation) | Frus. Cls. + Inv.Proj. · KITTI | RRE (°): 15.52\pm12.73 -> 7.56\pm7.63 | Helped |
| Add the cross-attention module (ablation) | DeepI2P · Oxford | RTE: 6.88 -> 1.65 | Helped |
| Add the cross-attention module (ablation) | DeepI2P · Oxford | RRE: 20.93 -> 4.14 | Helped |
| Reduce optimization initializations from 60 to 1 (ablation) | DeepI2P · Oxford | RTE: 1.65 -> 3.52 | Hurt |
| Reduce optimization initializations from 60 to 1 (ablation) | DeepI2P · Oxford | RRE: 4.14 -> 5.34 | Hurt |
| Reduce maximum translation limit from 10 m to 5 m (ablation) | DeepI2P · Oxford | RTE: 1.65 -> 1.52 | Helped |
| Increase maximum translation limit from 10 m to 15 m (ablation) | DeepI2P · Oxford | RTE: 1.65 -> 1.96 | Hurt |
| Reduce input point density from 20480 to 5120 points (ablation) | DeepI2P · Oxford | RTE: 1.65 -> 1.94 | Hurt |
Closest papers by meaning
- GraphI2P: Image-to-Point Cloud Registration with Exploring Pattern of Correspondence via Graph Learning (CVPR 2025 · similarity 0.90)
- FreeReg: Image-to-Point Cloud Registration Leveraging Pretrained Diffusion Models and Monocular Depth Estimators (ICLR 2024 · similarity 0.89)
- Implicit Correspondence Learning for Image-to-Point Cloud Registration (CVPR 2025 · similarity 0.88)
- Partial Point Cloud Registration with Multi-view 2D Image Learning (AAAI 2025 · similarity 0.87)
- 3D Focusing-and-Matching Network for Multi-Instance Point Cloud Registration (NeurIPS 2024 · similarity 0.87)
- A Consistency-Aware Spot-Guided Transformer for Versatile and Hierarchical Point Cloud Registration (NeurIPS 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.