Research map / 3D vision / Depth and stereo
COTR: Correspondence Transformer for Matching Across Images
Wei Jiang, Eduard Trulls, Jan Hosang, Andrea Tagliasacchi, Kwang Moo Yi
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 |
|---|---|---|---|
| replace transformer with MLP variant (ablation) | COTR · IMC2020 / MegaDepth warping | dense correspondence quality | Improved |
| add recursive zoom-in (four steps, factor 2) (ablation) | COTR · HPatches | pixel error distribution | Improved |
| add filtering of erroneous correspondences (ablation) | COTR · ETH3D | AEPE: roughly 5% relative | Improved |
| use linear positional-encoding frequency instead of log-linear (ablation) | COTR · MegaDepth training | training loss convergence | Improved |
| add dense interpolation instead of querying only sparse points | COTR · KITTI-2015 | AEPE: 2.62 -> 6.12 | Hurt |
| add dense interpolation | COTR · KITTI-2012 | Fl (%): 7.36 -> 10.50 | Hurt |
| add dense interpolation | COTR · HPatches | PCK-5px: 91.10 -> 86.33 | Hurt |
| increase number of matches per image pair N from 128 to 2048 | COTR+DEGENSAC · IMC2020 stereo | mAA(5deg): 0.356 -> 0.444 | Improved |
| COTR instead of RAFT with DEGENSAC | pose estimation · IMC2020 stereo | mAA(5deg): 0.163 -> 0.515 | Improved |
| COTR instead of RAFT | flow estimation · ETH3D (rate=15) | AEPE: 13.74 -> 2.61 | Improved |
| COTR instead of RAFT | flow estimation · KITTI-2015 | AEPE: 5.04 -> 2.62 | Improved |
| COTR instead of GLU-Net+GOCor | flow estimation · HPatches | AEPE: 20.16 -> 7.75 | Improved |
Later papers that compared with this method
10 later accepted papers report COTR as a baseline in their results tables; 8 of them report a better number on most of the results they share with it.
- FDR-Occ: Factorized Dense Routing for Full-Spectrum 3D Occupancy Prediction (ECCV 2026 · better on 3 of 3 shared results)
- Streaming Dense Voxel Representations for 3D Occupancy Prediction (ECCV 2026 · better on 0 of 2 shared results)
- Dr.Occ: Depth- and Region-Guided 3D Occupancy from Surround-View Cameras for Autonomous Driving (CVPR 2026 · better on 1 of 1 shared result)
- Robust 3D Semantic Occupancy Prediction With Calibration-Free Spatial Transformation (TPAMI 2026 · better on 1 of 1 shared result)
- ALOcc: Adaptive Lifting-Based 3D Semantic Occupancy and Cost Volume-Based Flow Predictions (ICCV 2025 · better on 4 of 4 shared results)
- STCOcc: Sparse Spatial-Temporal Cascade Renovation for 3D Occupancy and Scene Flow Prediction (CVPR 2025 · better on 2 of 2 shared results)
- Achieving Speed-Accuracy Balance in Vision-based 3D Occupancy Prediction via Geometric-Semantic Disentanglement (AAAI 2025 · better on 0 of 1 shared result)
- ProtoOcc: Accurate, Efficient 3D Occupancy Prediction Using Dual Branch Encoder-Prototype Query Decoder (AAAI 2025 · better on 1 of 1 shared result)
2 more not listed.
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
- Deep Matching Prior: Test-Time Optimization for Dense Correspondence (ICCV 2021 · similarity 0.90)
- Learning to Find Good Correspondences (CVPR 2018 · similarity 0.89)
- Universal Correspondence Network (NeurIPS 2016 · similarity 0.89)
- ETO:Efficient Transformer-based Local Feature Matching by Organizing Multiple Homography Hypotheses (NeurIPS 2024 · similarity 0.88)
- Diffusion Model for Dense Matching (ICLR 2024 · similarity 0.88)
- FC-GNN: Recovering Reliable and Accurate Correspondences from Interferences (CVPR 2024 · similarity 0.87)
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.
