Research map / Graph learning / Graph algorithms and clustering
Learning To Match Features With Seeded Graph Matching Network
Hongkai Chen, Zixin Luo, Jiahui Zhang, Lei Zhou, Xuyang Bai, Zeyu Hu, Chiew-Lan Tai, Long Quan
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 |
|---|---|---|---|
| select seeds by top-k scores instead of random (ablation) | SGMNet · YFCC100M (RootSIFT) | AUC@20: 71.25 -> 81.48 | Improved |
| add weighted unpooling (ablation) | SGMNet · YFCC100M (RootSIFT) | AUC@20: 78.64 -> 81.48 | Improved |
| add attentional pooling (ablation) | SGMNet · YFCC100M (RootSIFT) | AUC@20: 79.35 -> 81.48 | Improved |
| add reseeding (ablation) | SGMNet · YFCC100M (RootSIFT) | AUC@20: 80.41 -> 81.48 | Improved |
| add reseeding (ablation) | SGMNet · YFCC100M (RootSIFT) | precision: 84.47 -> 86.08 | Improved |
| learned seeding instead of ratio-score seeding (ablation) | SGMNet · YFCC100M (RootSIFT) | AUC@20: 81.48 -> 81.09 | No clear effect |
| learned seeding instead of ratio-score seeding (ablation) | SGMNet · YFCC100M (RootSIFT) | seed precision: 39.24 -> 51.22 | Improved |
| replace seed-based pooling with DiffPool (ablation) | SGMNet · YFCC100M (RootSIFT) | AUC@20: 81.48 -> 69.68 | Hurt |
| replace seed-based pooling with induced set attention (ablation) | SGMNet · YFCC100M (RootSIFT) | AUC@20: 81.48 -> 70.03 | Hurt |
| use SGMNet instead of SuperGlue | matcher · Aachen Day-Night night (RootSIFT), 0.25m 2deg | accuracy: 63.3 -> 70.4 | Improved |
| use SGMNet instead of SuperGlue with SuperPoint | matcher · YFCC100M (SuperPoint) | AUC@20: 80.00 -> 80.45 | No clear effect |
| use SGMNet instead of SuperGlue with RootSIFT | matcher · YFCC100M (RootSIFT) | AUC@20: 82.26 -> 81.48 | Hurt |
Closest papers by meaning
- SeedGNN: Graph Neural Network for Supervised Seeded Graph Matching (ICML 2023 · similarity 0.92)
- Combinatorial Learning of Robust Deep Graph Matching: An Embedding Based Approach (TPAMI 2020 · similarity 0.88)
- ClusterGNN: Cluster-Based Coarse-To-Fine Graph Neural Network for Efficient Feature Matching (CVPR 2022 · similarity 0.88)
- ResMatch: Residual Attention Learning for Feature Matching (AAAI 2024 · similarity 0.86)
- Discrete Cycle-Consistency Based Unsupervised Deep Graph Matching (AAAI 2024 · similarity 0.85)
- Learning Partial Graph Matching via Optimal Partial Transport (ICLR 2025 · 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.
