Research map / Point clouds / Registration and matching
MinCD-PnP: Learning 2D-3D Correspondences with Approximate Blind PnP
Pei An, Jiaqi Yang, Muyao Peng, You Yang, Qiong Liu, Xiaolin Wu, 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 L_Chamfer loss (ablation) | MATR+MinCD-Net · 7-Scenes Office-trained | IR: 0.489 -> 0.567 | Improved |
| add L_Chamfer loss (ablation) | MATR+MinCD-Net · 7-Scenes Office-trained | RR: 0.516 -> 0.646 | Improved |
| add keypoint loss L_key (ablation) | MATR+MinCD-Net · 7-Scenes Office-trained | IR: 0.473 -> 0.489 | Improved |
| add keypoint loss L_key (ablation) | MATR+MinCD-Net · 7-Scenes Office-trained | RR: 0.502 -> 0.516 | Improved |
| use Shi-Tomasi instead of uniform sampling for 2D keypoints (ablation) | MinCD-Net · 7-Scenes Office-trained | IR: 0.542 -> 0.567 | No clear effect |
| use SIFT instead of Shi-Tomasi keypoints (ablation) | MinCD-Net · 7-Scenes Office-trained | RR: 0.646 -> 0.638 | No clear effect |
| lower threshold s_th from e^-0.4 to e^-0.5 (ablation) | MinCD-Net · 3D keypoint precision | precision: 0.442 -> 0.308 | Hurt |
| raise s_th from e^-0.2 to e^-0.4 (ablation) | MinCD-Net · 3D keypoint recall | recall: 0.454 -> 0.722 | Improved |
| add MinCD-Net supervision to MATR | MATR · 7-Scenes trained on Office | RR: 0.491 -> 0.647 | Improved |
| add MinCD-Net supervision to MATR | MATR · 7-Scenes trained on Chess | IR: 0.387 -> 0.445 | Improved |
| add Diff. PnP supervision to MATR | MATR · 7-Scenes trained on Office | RR: 0.491 -> 0.501 | No clear effect |
| add Diff. PnP supervision to MATR | MATR · Efficiency table RR | RR: 51.0% -> 49.1% | Hurt |
Closest papers by meaning
- Solving the Blind Perspective-n-Point Problem End-To-End With Robust Differentiable Geometric Optimization (ECCV 2020 · similarity 0.89)
- A Practical O(N2) Outlier Removal Method for Point Cloud Registration (TPAMI 2021 · similarity 0.86)
- RPM-Net: Robust Point Matching Using Learned Features (CVPR 2020 · similarity 0.86)
- Deep Closest Point: Learning Representations for Point Cloud Registration (ICCV 2019 · similarity 0.86)
- Partial Point Cloud Registration with Multi-view 2D Image Learning (AAAI 2025 · similarity 0.85)
- Implicit Correspondence Learning for Image-to-Point Cloud Registration (CVPR 2025 · similarity 0.85)
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.
