Research map / Label-efficient and robust learning / Recognition and long-tail
BroadFace: Looking at Tens of Thousands of People at Once for Face Recognition
Yonghyun Kim, Wonpyo Park, Jongju Shin
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 compensation of enqueued embeddings at queue size 32768 (ablation) | BroadFace · IJB-B FAR=1e-6 | TAR: 0.4259 -> 0.4653 | Improved |
| add compensation at queue size 8192 (ablation) | BroadFace · IJB-B FAR=1e-6 | TAR: 0.4346 -> 0.4394 | Improved |
| add compensation at queue size 2048 (ablation) | BroadFace · IJB-B FAR=1e-6 | TAR: 0.4310 -> 0.4255 | Hurt |
| add compensation when queue exceeds 16k (ablation) | BroadFace · image retrieval recall | Recall | Improved |
| enlarge the queue from 0 to 2048 (ablation) | ArcFace · IJB-B FAR=1e-6 | TAR: 0.3828 -> 0.4255 | Improved |
| enlarge the queue from 2048 to 32768 (with compensation) (ablation) | BroadFace · IJB-B FAR=1e-6 | TAR: 0.4255 -> 0.4653 | Improved |
| add BroadFace to ArcFace | ArcFace · IJB-B FAR=1e-6 | TAR: 0.3828 -> 0.4653 | Improved |
| add BroadFace to CosFace | CosFace · IJB-B FAR=1e-6 | TAR: 0.3649 -> 0.4092 | Improved |
| add BroadFace to ArcFace | ArcFace · IJB-C FAR=1e-4 | TAR: 0.9603 -> 0.9603 | No clear effect |
| add BroadFace to ArcFace on ResNet-34 | ResNet-34 · IJB-B FAR=1e-6 | TAR: 0.3981 -> 0.4325 | Improved |
| add BroadFace to ArcFace on MobileFaceNet | MobileFaceNet · IJB-B FAR=1e-5 | TAR: 0.8456 -> 0.8458 | No clear effect |
| add BroadFace to ArcFace on ResNet-100 | ResNet-100 · IJB-B FAR=1e-5 | TAR: 0.8933 -> 0.9081 | Improved |
Later papers that compared with this method
3 later accepted papers report BroadFace as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Pairwise Similarity Learning is SimPLE (ICCV 2023 · better on 0 of 1 shared result)
- AdaFace: Quality Adaptive Margin for Face Recognition (CVPR 2022 · better on 0 of 4 shared results)
- Variational Prototype Learning for Deep Face Recognition (CVPR 2021 · better on 4 of 4 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- ArcFace — ArcFace: Additive Angular Margin Loss for Deep Face Recognition (CVPR 2019) · better on 13 of 13 shared results
- CosFace — CosFace: Large Margin Cosine Loss for Deep Face Recognition (CVPR 2018) · better on 10 of 10 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
- WebFace260M: A Benchmark for Million-Scale Deep Face Recognition (TPAMI 2022 · similarity 0.89)
- Do We Really Need to Collect Millions of Faces for Effective Face Recognition? (ECCV 2016 · similarity 0.89)
- AdaptiveFace: Adaptive Margin and Sampling for Face Recognition (CVPR 2019 · similarity 0.89)
- RegularFace: Deep Face Recognition via Exclusive Regularization (CVPR 2019 · similarity 0.88)
- WebFace260M: A Benchmark Unveiling the Power of Million-Scale Deep Face Recognition (CVPR 2021 · similarity 0.88)
- Vec2Face: Scaling Face Dataset Generation with Loosely Constrained Vectors (ICLR 2025 · similarity 0.88)
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.
