AI Paper Atlas

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

Learning Spatial-Semantic Relationship for Facial Attribute Recognition With Limited Labeled Data

Ying Shu, Yan Yan, Si Chen, Jing-Hao Xue, Chunhua Shen, Hanzi Wang

CVPR 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Margin-based face recognition losses · 39 citations (OpenAlex)

Check 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
Pretrain with PRT, PST, and PCT jointly before fine-tuningResNet-50 for facial attribute recognition · CelebA, 0.2% labeledaccuracy: 82.57 -> 86.67Improved
Pretrain with PRT, PST, and PCT jointly before fine-tuningResNet-50 for facial attribute recognition · LFWA, 5% labeledaccuracy: 73.90 -> 78.68Improved
Add PRT to PST and PCT (ablation)SSPL · CelebA, 0.2% labeledaccuracy: 85.39 -> 86.67Improved
Add PRT to PST and PCT (ablation)SSPL · LFWA, 5% labeledaccuracy: 76.82 -> 78.68Improved
Add PST to PRT (ablation)SSPL · CelebA, 0.2% labeledaccuracy: 85.36 -> 85.96Improved
Add PST to PRT (ablation)SSPL · LFWA, 5% labeledaccuracy: 77.31 -> 78.08Improved
Add PCT to PRT and PST (ablation)SSPL · CelebA, 0.2% labeledaccuracy: 85.96 -> 86.67Improved
Add PCT to PRT and PST (ablation)SSPL · LFWA, 5% labeledaccuracy: 78.08 -> 78.68Improved
Use all three auxiliary tasks rather than PRT alone (ablation)SSPL · CelebA, 100% labeledaccuracy: 91.72 -> 91.77No clear effect
Use all three auxiliary tasks rather than PRT alone (ablation)SSPL · LFWA, 100% labeledaccuracy: 85.72 -> 86.53Improved
Use a self-supervised spatial-semantic pretraining method instead of the strongest listed baselineFacial attribute recognition · CelebA, 100% labeledaccuracy: 92.98 -> 91.77Hurt
Use a self-supervised spatial-semantic pretraining method instead of the strongest listed baselineFacial attribute recognition · LFWA, 100% labeledaccuracy: 86.56 -> 86.53No clear effect

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