Research map / Label-efficient and robust learning / Domain adaptation and generalization
Bipartite Mode Matching for Vision Training Set Search from a Hierarchical Data Server
Yue Yao, Ruining Yang, Tom Gedeon
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 balanced K-means instead of standard K-means for source clustering (ablation) | BMM · AlicePerson re-ID (5% IDs) | mAP (%): 26.08 -> 28.16 | Improved |
| replace greedy direct matching (with duplicates) by Hungarian one-to-one matching (ablation) | BMM · AlicePerson re-ID (5% IDs) | mAP (%): 20.14 -> 26.08 | Improved |
| replace direct matching with duplicates deleted by Hungarian matching (ablation) | BMM · AlicePerson re-ID (5% IDs) | mAP (%): 22.07 -> 26.08 | Improved |
| add hierarchical clustering of the data server alone, keeping greedy search (ablation) | BMM baseline · AlicePerson re-ID | mAP | No clear effect |
| add bipartite mode matching alone, keeping flat clustering (ablation) | BMM baseline · AlicePerson re-ID | mAP | No clear effect |
| combine hierarchical server with bipartite matching (ablation) | BMM baseline · AlicePerson re-ID | mAP | Improved |
| use hierarchical instead of flat clustering on source server (ablation) | BMM · Cluster-number analysis (Fig. 4) | FID and mAP | Improved |
| searched training set with BMM instead of random sampling at 5% IDs | IDE on AlicePerson · AlicePerson re-ID (5% IDs) | mAP (%): 14.49 -> 26.08 | Improved |
| BMM instead of SnP on ExDark detection | RetinaNet · ExDark vehicle detection (5% images) | mAP (%): 32.15 -> 34.83 | Improved |
| add MMT pseudo-labeling to the searched training set | BMM · Market re-ID | mAP (%): 26.08% -> 78.95% | Improved |
| BMM + MMT instead of greedy search + MMT | MMT · Market re-ID | mAP (%): 76.98 -> 78.95 | Improved |
| increase number of source and target clusters beyond small values (ablation) | BMM · Cluster-number analysis | FID and mAP: FID 9.8, mAP 83.0% at L = 20 | Improved |
Closest papers by meaning
- Large-Scale Training Data Search for Object Re-Identification (CVPR 2023 · similarity 0.86)
- Towards Inheritable Models for Open-Set Domain Adaptation (CVPR 2020 · similarity 0.85)
- SIGMA++: Improved Semantic-complete Graph Matching for Domain Adaptive Object Detection (TPAMI 2023 · similarity 0.85)
- SIGMA: Semantic-Complete Graph Matching for Domain Adaptive Object Detection (CVPR 2022 · similarity 0.85)
- Efficient and Consistent Adversarial Bipartite Matching (ICML 2018 · similarity 0.85)
- Dataset Distillation for Pre-Trained Self-Supervised Vision Models (NeurIPS 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.
