Research map / Label-efficient and robust learning / Recognition and long-tail
Label Error Correction and Generation through Label Relationships
Zijun Cui, Yong Zhang, Qiang Ji
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 |
|---|---|---|---|
| train on MAP-corrected labels instead of original noisy labels (LR) | logistic regression · CK+ AU recognition | mean F1-score: 0.785 -> 0.830 | Improved |
| train on MAP-corrected labels (LR) | logistic regression · BP4D AU recognition | mean F1-score: 0.657 -> 0.687 | Improved |
| train on corrected labels (LR, missing labels completed) | logistic regression · EmotionNet AU recognition | mean F1-score: 0.491 -> 0.524 | Improved |
| train on corrected labels (CNN) | 3-layer CNN · EmotionNet AU recognition | mean F1-score: 0.620 -> 0.626 | Improved |
| train on corrected labels | linear SVM · a-Pascal attribute prediction | G-Mean: 0.743 -> 0.753 | Improved |
| train on corrected labels | MLTSVM · CK+ AU recognition | mean F1-score: 0.691 -> 0.784 | Improved |
| generate labels for MMI from relationships learned on CK+ | SVM · MMI AU recognition | mean F1-score: 0.482 -> 0.532 | Improved |
| remove object-level (expression) labels from label relationships (AUs only) (ablation) | MAPLB with LR · CK+ AU recognition | mean F1-score: 0.830 -> 0.809 | Hurt |
| use AU-only relationships to correct labels vs original labels (ablation) | LR · CK+ AU recognition | mean F1-score: 0.785 -> 0.809 | Improved |
| train on corrected labels | LR · CK+ AU prediction uncertainty | mean entropy: 0.194 -> 0.110 | Improved |
| use corrected labels for AU-to-expression pipeline | LR · CK+ expression recognition | accuracy: 0.820 -> 0.885 | Improved |
Closest papers by meaning
- Knowledge Amalgamation for Multi-Label Classification via Label Dependency Transfer (AAAI 2023 · similarity 0.87)
- Multi-Label Learning From Single Positive Labels (CVPR 2021 · similarity 0.87)
- Predicting Label Distribution from Ternary Labels (NeurIPS 2024 · similarity 0.87)
- Rethinking Consistent Multi-Label Classification under Inexact Supervision (ICLR 2026 · similarity 0.87)
- Dual Relation Semi-Supervised Multi-Label Learning (AAAI 2020 · similarity 0.87)
- Multi-View Multi-Label Classification via View-Label Matching Selection (AAAI 2025 · similarity 0.86)
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.
