AI Paper Atlas

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

Towards Robust Face Recognition with Comprehensive Search

Manyuan Zhang, Guanglu Song, Yu Liu, Hongsheng Li

ECCV 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Fair and quality-aware face recognition · 4 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 data-cleaning search to loss and backbone search (ablation)loss+backbone search · IJB-B TAR@FAR=1e-5, MS1MV2TAR (%): 83.67 -> 85.66Improved
add data-cleaning search to loss and backbone search (ablation)loss+backbone search · IJB-C TAR@FAR=1e-5, MS1MV2TAR (%): 89.55 -> 90.45Improved
search data cleaning only instead of hand-crafted cleaning (ablation)baseline · IJB-C TAR@FAR=1e-5, MS1MV2TAR (%): 87.29 -> 88.47Improved
search loss scales and margins only (ablation)ArcFace baseline · IJB-C TAR@FAR=1e-5, MS1MV2TAR (%): 87.29 -> 88.88Improved
search backbone width and depth only (ablation)modified MobileNet · IJB-C TAR@FAR=1e-5, MS1MV2TAR (%): 87.29 -> 87.92Improved
search all three jointly instead of each alone (ablation)best single-aspect search · IJB-B TAR@FAR=1e-5, MS1MV2TAR (%): 83.26 -> 85.66Improved
joint search vs hand-crafted baselinebaseline · IJB-C TAR@FAR=1e-4, MS1MV2TAR (%): 92.61 -> 93.58Improved
joint search vs hand-crafted baselinebaseline · IJB-C TAR@FAR=1e-5, Glint360KTAR (%): 86.33 -> 90.06Improved
joint search vs hand-crafted baselinebaseline · IJB-B TAR@FAR=1e-5, Glint360KTAR (%): 79.83 -> 86.06Improved
use features from Glint360K model instead of MS1MV2 model for cleaning (ablation)data search · IJB-C TAR@FAR=1e-5TAR (%): 88.80 -> 88.84No clear effect
train with combination searched on Glint360K instead of on MS1MV2 (ablation)joint search · IJB-C TAR@FAR=1e-5, MS1MV2 trainTAR (%): 90.45 -> 89.41Hurt
transferred Glint360K-searched combination vs hand-crafted baselinebaseline · IJB-C TAR@FAR=1e-5, MS1MV2 trainTAR (%): 87.29 -> 89.41Improved

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