Research map / Video / Action recognition
Actionness Inconsistency-Guided Contrastive Learning for Weakly-Supervised Temporal Action Localization
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 AICL instead of RSKP | weakly-supervised temporal action localization · THUMOS14 test | mAP@IoU AVG (0.1:0.7): 45.1 -> 46.4 | Helped |
| Use AICL instead of CO2-Net | weakly-supervised temporal action localization · ActivityNet1.2 | mAP@IoU AVG: 26.4 -> 29.9 | Helped |
| Use AICL instead of RSKP | weakly-supervised temporal action localization · ActivityNet1.3 validation | mAP@IoU AVG: 25.0 -> 27.6 | Helped |
| Add class-agnostic loss Lca to action loss La (ablation) | AICL on THUMOS14 · THUMOS14 | mAP@IoU AVG: 31.9 -> 43.2 | Helped |
| Add contrastive loss Lc (ablation) | AICL with La and Lca · THUMOS14 | mAP@IoU AVG: 43.2 -> 44.8 | Helped |
| Add multimodal consistency loss Lmc (ablation) | AICL with La and Lca · THUMOS14 | mAP@IoU AVG: 43.2 -> 44.6 | Helped |
| Add both contrastive loss Lc and multimodal consistency loss Lmc (ablation) | AICL with La and Lca · THUMOS14 | mAP@IoU AVG: 43.2 -> 46.0 | Helped |
| Add action consistency loss Lac to the objective (ablation) | AICL with La, Lca, Lc, and Lmc · THUMOS14 | mAP@IoU AVG: 46.0 -> 46.4 | Helped |
| Use late rather than early fusion in both branches (ablation) | AICL RGB and optical-flow features · THUMOS14 | mAP@IoU AVG: 40.7 -> 46.4 | Helped |
| Use late fusion in the class-agnostic branch and early fusion in the class-aware branch (ablation) | AICL RGB and optical-flow features · THUMOS14 | mAP@IoU AVG: 41.6 -> 45.7 | Helped |
| Apply AICL rather than the La + Lca baseline | action-background predictions in the class-aware branch · THUMOS14 | action-background classification accuracy (Abin): 37.3% -> 60.3% | Helped |
| Apply AICL rather than the La + Lca baseline | action-background predictions in the class-agnostic branch · THUMOS14 | action-background classification accuracy (Abin ca): 65.7% -> 65.3% | No clear effect |
Closest papers by meaning
- Similar Modality Enhancement and Action Consistency Learning for Weakly Supervised Temporal Action Localization (AAAI 2025 · similarity 0.92)
- Fine-Grained Temporal Contrastive Learning for Weakly-Supervised Temporal Action Localization (CVPR 2022 · similarity 0.90)
- Weakly Supervised Action Selection Learning in Video (CVPR 2021 · similarity 0.89)
- Distilling Vision-Language Pre-Training To Collaborate With Weakly-Supervised Temporal Action Localization (CVPR 2023 · similarity 0.89)
- PivoTAL: Prior-Driven Supervision for Weakly-Supervised Temporal Action Localization (CVPR 2023 · similarity 0.89)
- Bridge the Gap: From Weak to Full Supervision for Temporal Action Localization with PseudoFormer (CVPR 2025 · similarity 0.88)
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.