Research map / Label-efficient and robust learning / Recognition and long-tail
BoundaryFace: A Mining Framework with Noise Label Self-Correction for Face Recognition
Shijie Wu, Xun Gong
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 boundary hard-sample mining to label self-correction (BoundaryFace vs BoundaryF1) (ablation) | BoundaryF1 · AgeDB, 20% closed-set noise | AgeDB accuracy: 94.02 -> 94.22 | Improved |
| add label self-correction (BoundaryF1) to ArcFace (ablation) | ArcFace · AgeDB, 20% closed-set noise | AgeDB accuracy: 93 -> 94.02 | Improved |
| add label self-correction (BoundaryF1) to ArcFace (ablation) | ArcFace · AgeDB, C20/O20 mixed noise | AgeDB accuracy: 89.55 -> 92.5 | Improved |
| use BoundaryFace instead of CurricularFace | CurricularFace · AgeDB, 20% closed-set noise | AgeDB accuracy: 91.88 -> 94.22 | Improved |
| use BoundaryFace instead of ArcFace | ArcFace · AgeDB, C30/O10 mixed noise | AgeDB accuracy: 89.93 -> 93.38 | Improved |
| use BoundaryFace instead of ArcFace | ArcFace · AgeDB, clean CASIA-WebFace | AgeDB accuracy: 94.05 -> 94.4 | No clear effect |
| add CurricularFace mining on noisy data | ArcFace · AgeDB, 20% closed-set noise | AgeDB accuracy: 93 -> 91.88 | Hurt |
| set lambda to pi instead of 2.5 (ablation) | BoundaryFace · SLLFW | SLLFW accuracy: 97.9 -> 98.12 | No clear effect |
| use margin 0.3 instead of 0.5 with mostly open-set noise (ablation) | ArcFace · AgeDB, C10/O30 | AgeDB accuracy: 89.02 -> 91.82 | Improved |
| use margin 0.5 instead of 0.3 with mostly closed-set noise (ablation) | ArcFace · AgeDB, C30/O10 | AgeDB accuracy: 88.5 -> 89.93 | Improved |
Later papers that compared with this method
1 later accepted paper reports BoundaryFace as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- RepFace: Refining Closed-Set Noise with Progressive Label Correction for Face Recognition (AAAI 2025 · better on 4 of 5 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.
- CurricularFace — CurricularFace: Adaptive Curriculum Learning Loss for Deep Face Recognition (CVPR 2020) · better on 6 of 6 shared results
- ArcFace — ArcFace: Additive Angular Margin Loss for Deep Face Recognition (CVPR 2019) · better on 6 of 6 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
- AdaptiveFace: Adaptive Margin and Sampling for Face Recognition (CVPR 2019 · similarity 0.89)
- RepFace: Refining Closed-Set Noise with Progressive Label Correction for Face Recognition (AAAI 2025 · similarity 0.89)
- AdaFace: Quality Adaptive Margin for Face Recognition (CVPR 2022 · similarity 0.88)
- Mis-Classified Vector Guided Softmax Loss for Face Recognition (AAAI 2020 · similarity 0.88)
- Co-Mining: Deep Face Recognition With Noisy Labels (ICCV 2019 · similarity 0.87)
- UniTSFace: Unified Threshold Integrated Sample-to-Sample Loss for Face Recognition (NeurIPS 2023 · similarity 0.87)
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.
