Research map / Label-efficient and robust learning / Recognition and long-tail
LVFace: Progressive Cluster Optimization for Large Vision Models in Face Recognition
Jinghan You, Shanglin Li, Yuanrui Sun, Jiangchuan Wei, Mingyu Guo, Chao Feng, Jiao Ran
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 PCO stage 1 (feature alignment with negative sub-sampling) (ablation) | ViT-L · MFR-Ongoing MR-All | accuracy: 97.27 -> 97.52 | Improved |
| add PCO stage 2 (centroid stabilization) (ablation) | ViT-L · MFR-Ongoing MR-All | accuracy: 97.52 -> 98.22 | Improved |
| add PCO stage 3 (boundary refinement without NCS) (ablation) | ViT-L · MFR-Ongoing MR-All | accuracy: 98.22 -> 98.49 | Improved |
| add PCO to ArcFace loss (ablation) | ViT-B · IJB-C TAR@FAR=1e-5 | TAR: 96.11 -> 96.68 | Improved |
| add PCO to CosFace loss (ablation) | ViT-B · IJB-C TAR@FAR=1e-5 | TAR: 96.15 -> 97.00 | Improved |
| use CosFace instead of ArcFace as base loss with PCO (ablation) | LVFace ViT-B · IJB-C TAR@FAR=1e-5 | TAR: 96.68 -> 97.00 | Improved |
| scale backbone from Tiny to Base on Glint360K (ablation) | LVFace · IJB-C TAR@FAR=1e-5 | TAR: 95.63 -> 97.00 | Improved |
| scale backbone from Base to Large on Glint360K (ablation) | LVFace · IJB-C TAR@FAR=1e-5 | TAR: 97.00 -> 97.02 | No clear effect |
| train ViT-L on WebFace42M instead of Glint360K (ablation) | LVFace-L · IJB-C TAR@FAR=1e-5 | TAR: 97.02 -> 97.25 | Improved |
| use LVFace-L versus TopoFR on Mask subset | LVFace · MFR-Ongoing Mask | accuracy: 93.96 -> 93.56 | Hurt |
| LVFace-L versus TransFace-L at IJB-C 1e-6 (ablation) | LVFace-L · IJB-C TAR@FAR=1e-6 (Glint360K) | TAR: 89.71 -> 89.51 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 4 of 4 on most shared results.
- TopoFR — TopoFR: A Closer Look at Topology Alignment on Face Recognition (NeurIPS 2024) · better on 2 of 3 shared results
- UniTSFace — UniTSFace: Unified Threshold Integrated Sample-to-Sample Loss for Face Recognition (NeurIPS 2023) · better on 3 of 3 shared results
- AdaFace — AdaFace: Quality Adaptive Margin for Face Recognition (CVPR 2022) · better on 2 of 2 shared results
- ArcFace — ArcFace: Additive Angular Margin Loss for Deep Face Recognition (CVPR 2019) · better on 1 of 1 shared result
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
- TransFace: Calibrating Transformer Training for Face Recognition from a Data-Centric Perspective (ICCV 2023 · similarity 0.88)
- Vec2Face: Scaling Face Dataset Generation with Loosely Constrained Vectors (ICLR 2025 · similarity 0.87)
- TransFace++: Rethinking the Face Recognition Paradigm With a Focus on Accuracy, Efficiency, and Security (TPAMI 2025 · similarity 0.87)
- ProgressFace: Scale-Aware Progressive Learning for Face Detection (ECCV 2020 · similarity 0.87)
- FaceMoE: Mixture of Experts for Low-Resolution Face Recognition (ECCV 2026 · similarity 0.87)
- Masked Face Recognition with Generative-to-Discriminative Representations (ICML 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.
