Research map / Label-efficient and robust learning / Recognition and long-tail
Domain Balancing: Face Recognition on Long-Tailed Domains
Dong Cao, Xiangyu Zhu, Xingyu Huang, Jianzhu Guo, Zhen Lei
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 |
|---|---|---|---|
| Apply Domain Balancing | ResNet50 trained on MS1MV2 · LFW | verification accuracy (%): 99.62% to 99.78% | Improved |
| Apply Domain Balancing | ResNet50 · LFW BLUFR, VR@FAR=0.01% | verification rate: 99.23 to 99.53 | Improved |
| Apply Domain Balancing | ResNet50 · CALFW | verification accuracy (%): 95.45 to 96.08 | Improved |
| Apply Domain Balancing | ResNet50 · CPLFW | verification accuracy (%): 92.08 to 92.63 | Improved |
| Apply Domain Balancing | ResNet50 · AgeDB | verification accuracy (%): 97.83 to 97.90 | Improved |
| Apply Domain Balancing | ResNet50 · MegaFace Challenge1 | Rank-1 identification accuracy (%): 95.53 to 96.35 | Improved |
| Add DBM to RBM (ablation) | ResNet18 trained on CASIA-Webface · Average of LFW, CALFW, CPLFW and AgeDB | average accuracy (%): 92.12 to 92.88 | Improved |
| Add a soft gate to the residual enhancement (ablation) | RBM · Average of LFW, CALFW, CPLFW and AgeDB | average accuracy (%): 91.25 to 92.12 | Improved |
| Add RBM with its soft gate (ablation) | Baseline face recognition model · Average of LFW, CALFW, CPLFW and AgeDB | average accuracy (%): 91.35 to 92.12 | Improved |
| Set DFI nearest-neighbor K to 0 (ablation) | Domain Balancing · LFW | verification accuracy (%): 99.3 to 98.8 | Hurt |
| Set DFI nearest-neighbor K to 1,000 instead of 100 (ablation) | Domain Balancing · LFW | verification accuracy (%): 99.3 to 99.1 | Hurt |
| Set DFI nearest-neighbor K to 3,000 instead of 100 (ablation) | Domain Balancing · LFW | verification accuracy (%): 99.3 to 99.2 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ArcFace — ArcFace: Additive Angular Margin Loss for Deep Face Recognition (CVPR 2019) · better on 8 of 8 shared results
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
- Rethinking Class-Balanced Methods for Long-Tailed Visual Recognition From a Domain Adaptation Perspective (CVPR 2020 · similarity 0.89)
- "On Multi-Domain Long-Tailed Recognition, Imbalanced Domain Generalization and Beyond" (ECCV 2022 · similarity 0.89)
- Domain-Conditioned Class Imbalance: Why Global Class Balance Fails Across Domains (NeurIPS 2026 · similarity 0.87)
- Long-tailed Recognition with Model Rebalancing (NeurIPS 2025 · similarity 0.85)
- Difficulty-aware Balancing Margin Loss for Long-tailed Recognition (AAAI 2025 · similarity 0.85)
- Exploring Weight Balancing on Long-Tailed Recognition Problem (ICLR 2024 · 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.
