Research map / Label-efficient and robust learning / Recognition and long-tail
Deep Graphical Feature Learning for the Feature Matching Problem
Zhen Zhang, Wee Sun Lee
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 |
|---|---|---|---|
| add global pooling layer (ablation) | CMPNN geometric feature net · Synthetic, with outliers | accuracy | Improved |
| add global pooling layer (ablation) | CMPNN geometric feature net · Synthetic, no outliers | accuracy | No clear effect |
| use a deeper network (repeat second CMPNN layer three times) instead of normal (ablation) | CMPNN geometric feature net · Synthetic (noise 0.05) | accuracy | No clear effect |
| use a deeper/wider network instead of normal (ablation) | CMPNN geometric feature net · Synthetic (noise 0.05) | GPU memory: 76 -> 103 MB deeper; 76 -> 297 MB wider | Hurt |
| use normal depth/width instead of shallower or narrower (ablation) | CMPNN geometric feature net · Synthetic (noise 0.05) | accuracy | Improved |
| use output dim 64 instead of 8 (ablation) | CMPNN geometric feature net · Synthetic (outliers + rotation) | accuracy | Improved |
| use output dim 512 instead of 4096 (ablation) | CMPNN geometric feature net · Synthetic (outliers + rotation) | accuracy | Improved |
| use output dim 512 instead of 64 (ablation) | CMPNN geometric feature net · Synthetic (outliers + rotation) | accuracy | No clear effect |
| learn local features with CMPNN instead of PointNet | PointNet baseline · Pascal-PF (original points) | matching accuracy: 69.7 -> 88.5 | Improved |
| learn local features with CMPNN instead of PointNet | PointNet baseline · Pascal-PF (random rotation) | matching accuracy: 41.5 -> 69.9 | Improved |
| apply random rotation to input points | Ours · Pascal-PF | matching accuracy: 88.5 -> 69.9 | Hurt |
| apply random rotation to input points | PointNet · Pascal-PF | matching accuracy: 69.7 -> 41.5 | Hurt |
Closest papers by meaning
- Deep Learning of Graph Matching (CVPR 2018 · similarity 0.89)
- Combinatorial Learning of Robust Deep Graph Matching: An Embedding Based Approach (TPAMI 2020 · similarity 0.88)
- Proxy Graph Matching with Proximal Matching Networks (AAAI 2021 · similarity 0.88)
- Ensemble Quadratic Assignment Network for Graph Matching (IJCV 2024 · similarity 0.87)
- ClusterGNN: Cluster-Based Coarse-To-Fine Graph Neural Network for Efficient Feature Matching (CVPR 2022 · similarity 0.86)
- Learnable Graph Matching: A Practical Paradigm for Data Association (TPAMI 2024 · 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.
