Research map / Label-efficient and robust learning / Recognition and long-tail
Hyperspherical Learning in Multi-Label Classification
Bo Ke, Yunquan Zhu, Mengtian Li, Xiujun Shu, Ruizhi Qiao, Bo Ren
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 hyperspherical learning (ablation) | BCE multi-label classifier · VOC12 (single positive labels) | mAP: 86.9 -> 89.1 | Improved |
| add hyperspherical learning (ablation) | BCE multi-label classifier · COCO (single positive labels) | mAP: 66.2 -> 70.4 | Improved |
| add hyperspherical learning (ablation) | BCE multi-label classifier · CUB (single positive labels) | mAP: 19.1 -> 20.6 | Improved |
| add hyperspherical learning (ablation) | BCE multi-label classifier · NUS (single positive labels) | mAP: 42.8 -> 46.5 | Improved |
| add hyperspherical learning (ablation) | BCE multi-label classifier · VOC12, COCO, CUB, NUS (full labels) | mAP: 90.2 -> 91.3 (VOC12), 77.6 -> 78.4 (COCO), 32.1 -> 33.5 (CUB), 52.6 -> 53.6 (NUS) | Improved |
| add adaptive scale and margin learning (ablation) | HML · CUB (single positive labels) | mAP: 20.6 -> 21.1 | Improved |
| add adaptive scale and margin learning (ablation) | HML · VOC12, COCO, NUS (single positive labels) | mAP: 89.1 -> 89.1 (VOC12), 70.4 -> 70.7 (COCO), 46.5 -> 46.7 (NUS) | No clear effect |
| add adaptive scale and margin learning (ablation) | HML · NUS (full labels) | mAP: 53.6 -> 54.1 | Improved |
| add adaptive scale and margin learning (ablation) | HML · VOC12 and COCO (full labels) | mAP: 91.3 -> 91.3 (VOC12), 78.4 -> 78.6 (COCO) | No clear effect |
| apply angular margin to positives only (train-test inconsistency) (ablation) | HML · VOC12 (single positive labels) | mAP: 88.0 -> 83.4 | Hurt |
| apply angular margin to negatives only (ablation) | HML · VOC12 (single positive labels) | mAP: 88.0 -> 88.2 | No clear effect |
| apply angular margin to both positives and negatives (ablation) | HML · VOC12 (single positive labels) | mAP: 88.0 -> 89.1 | Improved |
Closest papers by meaning
- Learning with Noisy Labels Using Hyperspherical Margin Weighting (AAAI 2024 · similarity 0.88)
- In Pursuit of Causal Label Correlations for Multi-label Image Recognition (NeurIPS 2024 · similarity 0.88)
- Multi-Label Supervised Contrastive Learning (AAAI 2024 · similarity 0.87)
- Hyperspherical Classification with Dynamic Label-to-Prototype Assignment (CVPR 2024 · similarity 0.87)
- Semi-supervised Multi-label Learning with Balanced Binary Angular Margin Loss (NeurIPS 2024 · similarity 0.87)
- Addressing Multi-Label Learning with Partial Labels: From Sample Selection to Label Selection (AAAI 2025 · 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.
