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
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Pretrain with PRT, PST, and PCT jointly before fine-tuning | ResNet-50 for facial attribute recognition · CelebA, 0.2% labeled | accuracy: 82.57 -> 86.67 | Improved |
| Pretrain with PRT, PST, and PCT jointly before fine-tuning | ResNet-50 for facial attribute recognition · LFWA, 5% labeled | accuracy: 73.90 -> 78.68 | Improved |
| Add PRT to PST and PCT (ablation) | SSPL · CelebA, 0.2% labeled | accuracy: 85.39 -> 86.67 | Improved |
| Add PRT to PST and PCT (ablation) | SSPL · LFWA, 5% labeled | accuracy: 76.82 -> 78.68 | Improved |
| Add PST to PRT (ablation) | SSPL · CelebA, 0.2% labeled | accuracy: 85.36 -> 85.96 | Improved |
| Add PST to PRT (ablation) | SSPL · LFWA, 5% labeled | accuracy: 77.31 -> 78.08 | Improved |
| Add PCT to PRT and PST (ablation) | SSPL · CelebA, 0.2% labeled | accuracy: 85.96 -> 86.67 | Improved |
| Add PCT to PRT and PST (ablation) | SSPL · LFWA, 5% labeled | accuracy: 78.08 -> 78.68 | Improved |
| Use all three auxiliary tasks rather than PRT alone (ablation) | SSPL · CelebA, 100% labeled | accuracy: 91.72 -> 91.77 | No clear effect |
| Use all three auxiliary tasks rather than PRT alone (ablation) | SSPL · LFWA, 100% labeled | accuracy: 85.72 -> 86.53 | Improved |
| Use a self-supervised spatial-semantic pretraining method instead of the strongest listed baseline | Facial attribute recognition · CelebA, 100% labeled | accuracy: 92.98 -> 91.77 | Hurt |
| Use a self-supervised spatial-semantic pretraining method instead of the strongest listed baseline | Facial attribute recognition · LFWA, 100% labeled | accuracy: 86.56 -> 86.53 | No clear effect |
Closest papers by meaning
- SPL-Net: Spatial-Semantic Patch Learning Network for Facial Attribute Recognition with Limited Labeled Data (IJCV 2023 · similarity 0.93)
- Improving Facial Attribute Prediction Using Semantic Segmentation (CVPR 2017 · similarity 0.89)
- Self-Supervised Facial Representation Learning with Facial Region Awareness (CVPR 2024 · similarity 0.87)
- LAFS: Landmark-based Facial Self-supervised Learning for Face Recognition (CVPR 2024 · similarity 0.83)
- Attributes Shape the Embedding Space of Face Recognition Models (ICML 2025 · similarity 0.83)
- SSA-Seg: Semantic and Spatial Adaptive Pixel-level Classifier for Semantic Segmentation (NeurIPS 2024 · similarity 0.83)
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.
