AI Paper Atlas

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

SphereFace Revived: Unifying Hyperspherical Face Recognition

Weiyang Liu, Yandong Wen, Bhiksha Raj, Rita Singh, Adrian Weller

TPAMI 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Margin-based face recognition losses · 56 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 CGD (gradient detaching) with hard feature normalization (ablation)SphereFace · VGGFace2 validationAUC-0.0005: 46.02 -> 61.37Improved
add CGD (gradient detaching) with hard feature normalization (ablation)SphereFace-R v1 · VGGFace2 validationAUC-0.0005: 28.78 -> 60.45Improved
add CGD (gradient detaching) with hard feature normalization (ablation)SphereFace-R v2 · VGGFace2 validationAUC-0.0005: 36.61 -> 62.72Improved
add CGD with no feature normalization (ablation)SphereFace-R v1 · VGGFace2 validationAUC-0.0005: 49.92 -> 55.97Improved
add CGD with soft feature normalization (ablation)SphereFace-R v1 · VGGFace2 validationAUC-0.0005: 25.26 -> 61.37Improved
increase margin m from 1.2 to 1.5 without feature normalization (ablation)SphereFace-R v1 · VGGFace2 validationAUC-0.0005: 55.97 -> 31.54Hurt
use margin m=1.2 vs m=1.1 without feature normalization (ablation)SphereFace-R v2 · VGGFace2 validationAUC-0.0005: 52.71 -> 56.08Improved
soft feature normalization t=0.5 vs t=0.05 (ablation)SphereFace-R v1 · VGGFace2 validationAUC-0.0005: 39.43 -> 61.37Improved
soft instead of hard feature normalizationSphereFace-R v1 · IJB-C TAR@FAR 1e-4TAR: 89.67 -> 91.59Improved
soft instead of hard feature normalizationSphereFace-R v1 · MegaFace identificationaccuracy: 83.39 -> 89.17Improved
SphereFace-R v1 vs original SphereFace with SFNSphereFace-R v1 · IJB-B TAR@FAR 1e-4TAR: 87.63 -> 88.88Improved
SphereFace-R v1 vs original SphereFace with no feature normalizationSphereFace-R v1 · MS-Celeb-1MAUC-0.0005: 49.64 -> 84.66Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most 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

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