Research map / Label-efficient and robust learning / Recognition and long-tail
Active Self-Paced Learning for Cost-Effective and Progressive Face Identification
Liang Lin, Keze Wang, Deyu Meng, Wangmeng Zuo, Lei Zhang
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 changed | Where | Measured | Effect |
|---|---|---|---|
| combine active learning with self-paced learning (ASPL) instead of AL_RAND/CCAL/CPAL | incremental face identification · CACD | annotation amount to match accuracy: 81% -> 40% | Improved |
| combine active learning with self-paced learning (ASPL) instead of AL_RAND/CCAL/CPAL | incremental face identification · CASIA-WebFace-Sub | annotation amount to match accuracy: 65% -> 45% | Improved |
| add CNN fine-tuning (ablation) | ASPL · CASIA-WebFace-Sub | recognition accuracy | Improved |
| add self-paced learning (SPL) component (ablation) | ASPL · CASIA-WebFace-Sub | recognition accuracy | Improved |
| add active learning (AL) component (ablation) | ASPL · CASIA-WebFace-Sub | recognition accuracy | Improved |
| add CNN fine-tuning (ablation) | ASPL · CASIA-WebFace-Sub, iteration 5 | pseudo-label error rate: 8.2% -> 4.5% | Improved |
| add CNN fine-tuning (ablation) | ASPL · CASIA-WebFace-Sub, iteration 25 | pseudo-label error rate: 4.9% -> 3.3% | Improved |
| add active learning to self-paced learning (ablation) | ASPL · CASIA-WebFace-Sub (20 individuals) | sensitivity to initial samples | Improved |
| handle unseen new classes incrementally instead of giving all classes at start (ablation) | ASPL · CASIA-WebFace-Sub, 600 classes | accuracy: 81.0% -> 81.6% | No clear effect |
| handle unseen new classes incrementally instead of giving all classes at start (ablation) | ASPL · CASIA-WebFace-Sub, 925 classes | accuracy: 76.0% -> 76.0% | No clear effect |
| add annotated-sample verification step (ablation) | ASPL · CASIA-WebFace-Sub with noisy labels | recovery speed from noise | Improved |
| add noisy initial annotations (10-50%) | ASPL · CASIA-WebFace-Sub | test accuracy | Hurt |
Closest papers by meaning
- Accelerated Training for Massive Classification via Dynamic Class Selection (AAAI 2018 · similarity 0.87)
- BroadFace: Looking at Tens of Thousands of People at Once for Face Recognition (ECCV 2020 · similarity 0.86)
- 50 Years of Automated Face Recognition (TPAMI 2026 · similarity 0.86)
- How to Boost Face Recognition with StyleGAN? (ICCV 2023 · similarity 0.86)
- Vec2Face: Scaling Face Dataset Generation with Loosely Constrained Vectors (ICLR 2025 · similarity 0.86)
- UIFace: Unleashing Inherent Model Capabilities to Enhance Intra-Class Diversity in Synthetic Face Recognition (ICLR 2025 · similarity 0.85)
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.
