Research map / Label-efficient and robust learning / Semi-supervised learning
Improving Weakly Supervised Temporal Action Localization by Bridging Train-Test Gap in Pseudo Labels
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use boundary-derived pseudo-labels with Gaussian fusion, LinPro generation, and Δ pseudo labels | RSKP · THUMOS14 | AVG mAP (0.1:0.7): 46.4 -> 48.3 | Helped |
| Use the proposed pseudo-label pipeline | RSKP · ActivityNet1.3 validation | AVG mAP: 25.1 -> 28.8 | Helped |
| Replace NMS with Gaussian weighted fusion during training (ablation) | RSKP · THUMOS14 training | AVG mAP: 44.9 -> 45.6 | Helped |
| Use uniform rather than Gaussian fusion (ablation) | RSKP · THUMOS14 training | AVG mAP: 44.9 -> 31.2 | Hurt |
| Use exponential rather than Gaussian fusion (ablation) | RSKP · THUMOS14 training | AVG mAP: 45.3 vs 45.6 | No clear effect |
| Use t-distribution rather than Gaussian fusion (ablation) | RSKP · THUMOS14 training | AVG mAP: 45.6 vs 45.6 | No clear effect |
| Replace NMS with Gaussian weighted fusion during testing (ablation) | RSKP · THUMOS14 testing | AVG mAP: 44.9 -> 47.1 | Helped |
| Use uniform rather than Gaussian fusion (ablation) | RSKP · THUMOS14 testing | AVG mAP: 44.9 -> 41.5 | Hurt |
| Use exponential rather than Gaussian fusion (ablation) | RSKP · THUMOS14 testing | AVG mAP: 47.1 vs 47.1 | No clear effect |
| Use t-distribution rather than Gaussian fusion (ablation) | RSKP · THUMOS14 testing | AVG mAP: 47.1 vs 47.1 | No clear effect |
| Add LinPro pseudo-label generation to NMS-generated boundaries (ablation) | RSKP · THUMOS14 | AVG mAP: 45.1 -> 44.9 | Hurt |
| Add Gaussian weighted fusion after LinPro pseudo-label generation (ablation) | RSKP · THUMOS14 | AVG mAP: 44.9 -> 45.6 | Helped |
Closest papers by meaning
- Bridge the Gap: From Weak to Full Supervision for Temporal Action Localization with PseudoFormer (CVPR 2025 · similarity 0.95)
- Rethinking Pseudo-Label Guided Learning for Weakly Supervised Temporal Action Localization from the Perspective of Noise Correction (AAAI 2025 · similarity 0.94)
- Weakly-Supervised Temporal Action Localization by Inferring Salient Snippet-Feature (AAAI 2024 · similarity 0.90)
- HR-Pro: Point-Supervised Temporal Action Localization via Hierarchical Reliability Propagation (AAAI 2024 · similarity 0.90)
- Similar Modality Enhancement and Action Consistency Learning for Weakly Supervised Temporal Action Localization (AAAI 2025 · similarity 0.90)
- Realigning Confidence with Temporal Saliency Information for Point-Level Weakly-Supervised Temporal Action Localization (CVPR 2024 · similarity 0.89)
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.