Research map / Point clouds / Point cloud classification
Bootstrap Your Own Correspondences
Mohamed El Banani, Justin Johnson
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 visual bootstrapping (visual correspondences) to geometric training (ablation) | BYOC · ScanNet registration | rotation accuracy at 5 degrees (%): 80.3 -> 86.5 | Improved |
| add visual bootstrapping (visual correspondences) (ablation) | BYOC · 3D Match FMR | feature-match recall: 0.786 -> 0.766 | Hurt |
| add rotation augmentation (ablation) | BYOC · 3D Match FMR | feature-match recall: 0.766 -> 0.827 | Improved |
| add rotation augmentation (ablation) | BYOC · ScanNet registration | rotation accuracy at 5 degrees (%): 86.5 -> 71.2 | Hurt |
| train on ScanNet instead of 3D Match (ablation) | BYOC · ScanNet registration | rotation accuracy at 5 degrees (%): 66.5 -> 86.5 | Improved |
| train on ScanNet instead of 3D Match (ablation) | BYOC · 3D Match FMR | feature-match recall: 0.690 -> 0.766 | Improved |
| use Lowe-ratio filtered correspondences instead of all features (ablation) | BYOC · 3D Match FMR | feature-match recall: 0.662 -> 0.766 | Improved |
| use Weighted Procrustes with Lowe ratio instead of RANSAC (ablation) | BYOC · ScanNet registration | rotation accuracy at 5 degrees (%): 81.3 -> 86.5 | Improved |
| use RANSAC instead of Weighted Procrustes (ablation) | FCGF · ScanNet registration | rotation accuracy at 5 degrees (%): 54.1 -> 75.3 | Improved |
| use visual features instead of geometric features for registration (ablation) | BYOC · ScanNet registration | rotation mean error (degrees): 3.8 -> 2.7 | Improved |
| use random visual features instead of random geometric features (ablation) | untrained encoders · ScanNet registration | rotation mean error (degrees): 21.3 -> 6.4 | Improved |
| replace pose-supervised FCGF+RANSAC by unsupervised BYOC | registration pipeline · ScanNet registration | rotation accuracy at 5 degrees (%): 75.3 -> 86.5 | Improved |
Closest papers by meaning
- Improving RGB-D Point Cloud Registration by Learning Multi-Scale Local Linear Transformation (ECCV 2022 · similarity 0.88)
- Learning general and distinctive 3D local deep descriptors for point cloud registration (TPAMI 2022 · similarity 0.88)
- Mining and Transferring Feature-Geometry Coherence for Unsupervised Point Cloud Registration (NeurIPS 2024 · similarity 0.88)
- Learning Compact Geometric Features (ICCV 2017 · similarity 0.87)
- ToThePoint: Efficient Contrastive Learning of 3D Point Clouds via Recycling (CVPR 2023 · similarity 0.87)
- Sonata: Self-Supervised Learning of Reliable Point Representations (CVPR 2025 · 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.
