Research map / 3D vision / Depth and stereo
ConvMatch: Rethinking Network Design for Two-View Correspondence Learning
Shihua Zhang, Jiayi Ma
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 |
|---|---|---|---|
| Use ConvMatch's motion-field regularization and CNN rectification framework | Two-view correspondence learning · YFCC100M outdoor relative pose, with RANSAC | AUC@5°: 31.55 -> 31.69 | Improved |
| Use ConvMatch's motion-field regularization and CNN rectification framework | Two-view correspondence learning · YFCC100M outdoor relative pose, without RANSAC | AUC@10°: 44.69 -> 49.14 | Improved |
| Use ConvMatch's motion-field regularization and CNN rectification framework | Two-view correspondence learning · SUN3D indoor relative pose, with RANSAC | AUC@20°: 33.47 -> 34.41 | Improved |
| Use ConvMatch's motion-field regularization and CNN rectification framework | Two-view correspondence learning · SUN3D indoor relative pose, without RANSAC | AUC@20°: 37.15 -> 40.49 | Improved |
| Use ConvMatch for homography estimation | Two-view correspondence learning · HPatches homography estimation | Acc.: 72.93 -> 73.45 | Improved |
| Use ConvMatch for homography estimation | Two-view correspondence learning · HPatches homography estimation | F-score: 85.44 -> 83.80 | Hurt |
| Classify unordered correspondences directly rather than regularizing and rectifying motion vectors (ablation) | ConvMatch · YFCC100M outdoor relative pose, with RANSAC | AUC@20°: 68.45 -> 65.92 | Hurt |
| Replace the CNN backbone with an MLP and Context Normalization (ablation) | ConvMatch · YFCC100M outdoor relative pose, with RANSAC | AUC@20°: 68.45 -> 66.21 | Hurt |
| Replace the simple CNN backbone with UNet (ablation) | ConvMatch · YFCC100M outdoor relative pose, with RANSAC | AUC@20°: 68.45 -> 69.96 | Improved |
| Increase the number of rectifying layers from L = 6 to L = 8 (ablation) | ConvMatch · YFCC100M outdoor relative pose, without RANSAC | AUC@20°: 67.91 -> 68.89 | Improved |
| Increase the grid number from K = 16 to K = 24 (ablation) | ConvMatch · YFCC100M outdoor relative pose, without RANSAC | AUC@20°: 67.91 -> 67.94 | No clear effect |
Later papers that compared with this method
3 later accepted papers report ConvMatch+ as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- SC-Net: Robust Correspondence Learning via Spatial and Cross-Channel Context (AAAI 2026 · better on 2 of 2 shared results)
- Boosting Correspondence Learning with Structure-Aware Estimator (ECCV 2026 · better on 1 of 2 shared results)
- BCLNet: Bilateral Consensus Learning for Two-View Correspondence Pruning (AAAI 2024 · better on 0 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- SuperGlue — SuperGlue: Learning Feature Matching With Graph Neural Networks (CVPR 2020) · better on 0 of 1 shared result
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
- Universal Correspondence Network (NeurIPS 2016 · similarity 0.88)
- DeMatch++: Two-View Correspondence Learning via Deep Motion Field Decomposition and Respective Local-Context Aggregation (TPAMI 2025 · similarity 0.88)
- SC-Net: Robust Correspondence Learning via Spatial and Cross-Channel Context (AAAI 2026 · similarity 0.87)
- DeMatch: Deep Decomposition of Motion Field for Two-View Correspondence Learning (CVPR 2024 · similarity 0.87)
- NCNet: Neighbourhood Consensus Networks for Estimating Image Correspondences (TPAMI 2020 · similarity 0.87)
- DeMo: Deep Motion Field Consensus with Learnable Kernels for Two-view Correspondence Learning (AAAI 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.
