Research map / Architectures and efficiency / RNNs and normalization
Learning to Find Good Correspondences
Read the paperCheck 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 learned correspondence filtering before RANSAC instead of stand-alone RANSAC | RANSAC pose estimation · Outdoors, SIFT | mAP@20°: 0.221 -> 0.462 | Helped |
| Use learned correspondence filtering before RANSAC instead of stand-alone RANSAC | RANSAC pose estimation · Outdoors, LIFT | mAP@20°: 0.291 -> 0.530 | Helped |
| Use learned correspondence filtering before RANSAC instead of stand-alone RANSAC | RANSAC pose estimation · Indoors, SIFT | mAP@20°: 0.097 -> 0.242 | Helped |
| Use learned correspondence filtering before RANSAC instead of stand-alone RANSAC | RANSAC pose estimation · Indoors, LIFT | mAP@20°: 0.115 -> 0.222 | Helped |
| Add learned correspondence filtering, using the weighted 8-point estimate without test-time RANSAC | Stand-alone RANSAC · Outdoors, SIFT | mAP@20°: 0.221 -> 0.264 | Helped |
| Add learned correspondence filtering, using the weighted 8-point estimate without test-time RANSAC | Stand-alone RANSAC · Outdoors, LIFT | mAP@20°: 0.291 -> 0.343 | Helped |
| Add learned correspondence filtering, using the weighted 8-point estimate without test-time RANSAC | Stand-alone RANSAC · Indoors, SIFT | mAP@20°: 0.097 -> 0.148 | Helped |
| Add learned correspondence filtering, using the weighted 8-point estimate without test-time RANSAC | Stand-alone RANSAC · Indoors, LIFT | mAP@20°: 0.115 -> 0.143 | Helped |
| Add test-time RANSAC after learned correspondence filtering (ablation) | Ours + 8-point · Outdoors, SIFT | mAP@20°: 0.264 -> 0.462 | Helped |
| Add test-time RANSAC after learned correspondence filtering (ablation) | Ours + 8-point · Indoors, SIFT | mAP@20°: 0.148 -> 0.242 | Helped |
| Combine correspondence classification with essential-matrix regression (ablation) | Classification-only model · Sacre Coeur | mAP: 12-24% | Helped |
| Add essential-matrix regression after 20k batches of classification training (ablation) | Classification-only model · Training | relative performance: 5 to 20% | Helped |
Closest papers by meaning
- Zoom and Learn: Generalizing Deep Stereo Matching to Novel Domains (CVPR 2018 · similarity 0.87)
- Universal Correspondence Network (NeurIPS 2016 · similarity 0.87)
- Stereo Risk: A Continuous Modeling Approach to Stereo Matching (ICML 2024 · similarity 0.86)
- Robust Synthetic-to-Real Transfer for Stereo Matching (CVPR 2024 · similarity 0.86)
- Stereo Anywhere: Robust Zero-Shot Deep Stereo Matching Even Where Either Stereo or Mono Fail (CVPR 2025 · similarity 0.85)
- Hierarchical Object-Aware Dual-Level Contrastive Learning for Domain Generalized Stereo Matching (NeurIPS 2024 · 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.