Research map / Label-efficient and robust learning / Recognition and long-tail
WebFace260M: A Benchmark Unveiling the Power of Million-Scale Deep Face Recognition
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Train on WebFace42M instead of MS1MV2 | ArcFace, ResNet-100 · IJB-C | TAR@FAR=1e-4: 96.03 -> 97.70 | Helped |
| Increase training data from WebFace4M to WebFace12M to WebFace42M (ablation) | ArcFace, ResNet-100 · IJB-C | TAR@FAR=1e-4: 96.77 -> 97.47 -> 97.70 | Helped |
| Increase training data from WebFace4M to WebFace12M to WebFace42M (ablation) | ArcFace, ResNet-14 · IJB-C | TAR@FAR=1e-4: 93.13 -> 93.92 -> 94.22 | Helped |
| Use three CAST iterations rather than one or two (ablation) | CAST cleaning of WebFace260M · IJB-C | TAR@FAR=1e-4: 97.28 -> 97.51 -> 97.70 | Helped |
| Use a fourth CAST iteration rather than three (ablation) | CAST cleaning of WebFace260M · IJB-C | TAR@FAR=1e-4: 97.70 -> 97.66 | No clear effect |
| Use DBSCAN rather than K-Means for intra-class cleaning (ablation) | CAST cleaning on MS1M · IJB-C | TAR@FAR=1e-4: 96.03 -> 96.55 | Helped |
| Use DBSCAN rather than GCN-D for intra-class cleaning (ablation) | CAST cleaning on MS1M · IJB-C | TAR@FAR=1e-4: 96.48 -> 96.55 | Helped |
| Use DBSCAN rather than GCN-V for intra-class cleaning (ablation) | CAST cleaning on MS1M · IJB-C | TAR@FAR=1e-4: 96.42 -> 96.55 | Helped |
| Scale distributed training from one node to 32 nodes (ablation) | ResNet-100 + ArcFace on WebFace42M · IJB-C | training time: 233h -> 9h | Helped |
| Scale distributed training from one node to 32 nodes (ablation) | ResNet-100 + ArcFace on WebFace42M · IJB-C | TAR@FAR=1e-4: 97.71 -> 97.70 | No clear effect |
Closest papers by meaning
- The MegaFace Benchmark: 1 Million Faces for Recognition at Scale (CVPR 2016 · similarity 0.89)
- Vec2Face: Scaling Face Dataset Generation with Loosely Constrained Vectors (ICLR 2025 · similarity 0.87)
- HyperFace: Generating Synthetic Face Recognition Datasets by Exploring Face Embedding Hypersphere (ICLR 2025 · similarity 0.87)
- FSFM: A Generalizable Face Security Foundation Model via Self-Supervised Facial Representation Learning (CVPR 2025 · similarity 0.86)
- UIFace: Unleashing Inherent Model Capabilities to Enhance Intra-Class Diversity in Synthetic Face Recognition (ICLR 2025 · similarity 0.85)
- GIF: Generative Inspiration for Face Recognition at Scale (CVPR 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.