AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Caption-Supervised Face Recognition: Training a State-of-the-Art Face Model without Manual Annotation

Qingqiu Huang, Lei Yang, Huaiyi Huang, Tong Wu, Dahua Lin

ECCV 2020 · Label-efficient and robust learning › Recognition and long-tail · approach: Fair and quality-aware face recognition · 26 citations (OpenAlex)

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 changedWhereMeasuredEffect
add propagation data (S2) to One2One data (ablation)CSFR · MegaFacetop-1 identification (%): 79.81 -> 82.41Improved
add expansion data (S3) to One2One plus propagation data (ablation)CSFR · MegaFacetop-1 identification (%): 82.41 -> 83.34Improved
add propagation data (S2) to One2One (ablation)CSFR · IJB-ATAR@FAR=0.001 (%): 68.67 -> 75.06Improved
add expansion data (S3) to One2One plus propagation (ablation)CSFR · IJB-ATAR@FAR=0.001 (%): 75.06 -> 77.50Improved
train on automatically annotated data only (S2+S3), dropping One2One (ablation)CSFR · MegaFacetop-1 identification (%): 69.52 vs 79.81No clear effect
add K-means clustering of unlabeled faces to One2OneS1 baseline · MegaFacetop-1 identification (%): 79.81 -> 79.9No clear effect
add LTC clustering of unlabeled faces to One2OneS1 baseline · IJB-ATAR@FAR=0.001 (%): 68.67 -> 71.34Improved
use max-pooling instead of mean-pooling in MIL (ablation)MIL · MegaFacetop-1 identification (%): 69.62 -> 73.11Improved
use CSFR instead of best MIL method (max-pooling)CSFR · IJB-ATAR@FAR=0.001 (%): 49.06 -> 77.50Improved
train ArcFace on MovieFace instead of MS1MArcFace · IJB-ATAR@FAR=0.001 (%): 96.82 -> 96.96Improved
train ArcFace on MovieFace instead of MS1MArcFace · MegaFacetop-1 identification (%): 97.92 -> 96.96Hurt
train ArcFace on MovieFace instead of IMDb-FaceArcFace · IJB-ATAR@FAR=0.001 (%): 94.25 -> 96.96Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy