Research map / Graph learning / Graph algorithms and clustering
ClusterGNN: Cluster-Based Coarse-To-Fine Graph Neural Network for Efficient Feature Matching
Check 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 coarse-to-fine increasing cluster counts rather than a fixed count (ablation) | ClusterGNN · YFCC100M, ASLFeat, 2K keypoints | pose AUC at 5°: 31.26 / 37.05 / 37.44 / 36.77 -> 42.62 | Helped |
| replace dual softmax with Sinkhorn (ablation) | ClusterGNN · YFCC100M, ASLFeat, 2K keypoints | pose AUC at 5°: 42.62 -> 38.83 | Hurt |
| replace dual softmax with Sinkhorn (ablation) | ClusterGNN · YFCC100M, ASLFeat, 2K keypoints | runtime: 61 -> 68 | Hurt |
| replace dual softmax with Sinkhorn (ablation) | ClusterGNN · YFCC100M, ASLFeat, 2K keypoints | GPU memory: 94 -> 94 | No clear effect |
| use ClusterGNN instead of SuperGlue | ASLFeat matching · YFCC100M | pose AUC at 5°: 37.50 -> 42.62 | Helped |
| use ClusterGNN instead of SuperGlue | ASLFeat matching · HPatches overall | homography accuracy (ε < 1/3/5 px): 0.49 / 0.83 / 0.89 -> 0.51 / 0.83 / 0.89 | Mixed |
| use ClusterGNN instead of SuperGlue | SuperPoint matching · HPatches overall | homography accuracy (ε < 1/3/5 px): 0.51 / 0.83 / 0.89 -> 0.52 / 0.84 / 0.90 | Mixed |
| use ClusterGNN instead of SuperGlue | SuperPoint matching · InLoc, DUC1 | localization accuracy at (0.5m,10°): 68.7 -> 69.7 | Helped |
| use ClusterGNN instead of SuperGlue | ASLFeat matching · InLoc, DUC1 | localization accuracy at (0.25m,10°): 51.5 -> 52.5 | Helped |
| use ClusterGNN instead of SuperGlue | SuperPoint matching · Aachen Day-Night, Day | localization accuracy at (0.5m,5°): 95.4 -> 95.5 | No clear effect |
| use ClusterGNN instead of SuperGlue | SuperPoint matching · Aachen Day-Night, Night | localization accuracy at (0.25m,2°): 86.7 -> 81.6 | Hurt |
| use ClusterGNN instead of SuperGlue | ASLFeat matching · Aachen Day-Night, Night | localization accuracy at (0.5m,5°): 91.8 -> 93.9 | Helped |
Closest papers by meaning
- Graph-Context Attention Networks for Size-Varied Deep Graph Matching (CVPR 2022 · similarity 0.88)
- ResMatch: Residual Attention Learning for Feature Matching (AAAI 2024 · similarity 0.88)
- LatentGNN: Learning Efficient Non-local Relations for Visual Recognition (ICML 2019 · similarity 0.88)
- Learning deep graph matching with channel-independent embedding and Hungarian attention (ICLR 2020 · similarity 0.87)
- FC-GNN: Recovering Reliable and Accurate Correspondences from Interferences (CVPR 2024 · similarity 0.85)
- MGNet: Learning Correspondences via Multiple Graphs (AAAI 2024 · similarity 0.84)
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.