Research map / Label-efficient and robust learning / Recognition and long-tail
Holistic Label Correction for Noisy Multi-Label Classification
Xiaobo Xia, Jiankang Deng, Wei Bao, Yuxuan Du, Bo Han, Shiguang Shan, Tongliang Liu
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 |
|---|---|---|---|
| add label dependence to the holistic score (not only instance-label dependence) (ablation) | HLC · Noisy Pascal-VOC 2007, Sym. 50% | mAP: 67.06 -> 68.00 | Improved |
| add label dependence to the holistic score (ablation) | HLC · Noisy Pascal-VOC 2012, Sym. 50% | mAP: 67.88 -> 68.69 | Improved |
| add label dependence to the holistic score (ablation) | HLC · Noisy MS-COCO, Sym. 50% | mAP: 46.21 -> 48.15 | Improved |
| add holistic label correction on top of ADDGCN (ablation) | ADDGCN · Noisy Pascal-VOC 2007, Sym. 50% | mAP: 44.14 -> 68.00 | Improved |
| vary correction threshold delta in 0.25-0.45 (ablation) | HLC · Noisy Pascal-VOC 2007 and 2012 | mAP | No clear effect |
| label correction (JOINT) vs no correction (APL, CDR) | baselines · Noisy Pascal-VOC 2007, Sym. 50% | mAP: JOINT 53.27 vs APL 47.64, CDR 49.12 | Improved |
| replace HLC with HLC+ASL loss | HLC · NUS-WIDE | mAP: 63.1 -> 64.5 | Improved |
| use HLC at small image size 112x112 | HLC · Noisy MS-COCO (112x112) | mAP: 35.98 vs CCMN 36.17 | No clear effect |
| multi-label methods designed for clean labels (CSRA, ADDGCN) versus plain BCE | noisy multi-label training · Noisy Pascal-VOC 2007, Sym. 40% | mAP: BCE 58.65 vs ADDGCN 55.75 | Hurt |
Closest papers by meaning
- Towards Reliable Multi-Label Classification via Conditional Dependency Modeling (ECCV 2026 · similarity 0.92)
- Learning with Feature-Dependent Label Noise: A Progressive Approach (ICLR 2021 · similarity 0.90)
- Limited-Supervised Multi-Label Learning with Dependency Noise (AAAI 2024 · similarity 0.89)
- Multi-Label Learning with Stronger Consistency Guarantees (NeurIPS 2024 · similarity 0.87)
- Addressing Multi-Label Learning with Partial Labels: From Sample Selection to Label Selection (AAAI 2025 · similarity 0.87)
- Adaptive Integration of Partial Label Learning and Negative Learning for Enhanced Noisy Label Learning (AAAI 2024 · similarity 0.87)
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.
