Research map / Video / Action recognition
Temporal Structure Mining for Weakly Supervised Action Detection
Tan Yu, Zhou Ren, Yuncheng Li, Enxu Yan, Ning Xu, Junsong Yuan
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 |
|---|---|---|---|
| use 2 phases instead of 1 (temporal structure) (ablation) | TSM · THUMOS14 | mAP@IoU avg: 21.6 -> 23.3 | Improved |
| use 3 phases instead of 1 (ablation) | TSM · THUMOS14 | mAP@IoU avg: 21.6 -> 23.4 | Improved |
| increase phases from 3 to 4 (ablation) | TSM · THUMOS14 | mAP@IoU avg: 23.4 -> 22.7 | Hurt |
| increase phases from 3 to 5 (ablation) | TSM · THUMOS14 | mAP@IoU avg: 23.4 -> 22.3 | Hurt |
| replace hard maximal path with soft-max path discovery (ablation) | TSM 2-phase two-stream · THUMOS14 | mAP@IoU0.3: 38.7 -> 39.3 | Improved |
| replace hard maximal path with soft-max path discovery (ablation) | TSM 2-phase two-stream · THUMOS14 | mAP@IoU0.5: 24.0 -> 24.6 | Improved |
| replace hard maximal path with soft-max path discovery (ablation) | TSM 2-phase two-stream · THUMOS14 | mAP@IoU0.7: 6.2 -> 6.6 | Improved |
| soft-max path discovery trains faster than hard (ablation) | TSM · THUMOS14 | training epochs: 45 -> 30 | Improved |
| add optical flow stream to RGB stream (ablation) | TSM 2-phase · THUMOS14 | mAP@IoU0.5: 15.1 -> 24.6 | Improved |
| add RGB stream to optical flow (fusion) (ablation) | TSM 2-phase · THUMOS14 | mAP@IoU0.7: 6.9 -> 6.6 | Hurt |
| add optical flow stream to RGB stream (ablation) | TSM 3-phase · ActivityNet 1.3 | mAP@IoU0.5: 28.1 -> 30.3 | Improved |
| add RGB stream to optical flow (ablation) | TSM 3-phase · ActivityNet 1.3 | mAP@IoU0.5: 23.1 -> 30.3 | Improved |
Closest papers by meaning
- Action Sets: Weakly Supervised Action Segmentation Without Ordering Constraints (CVPR 2018 · similarity 0.90)
- Cross-Task Relation-Aware Consistency for Weakly Supervised Temporal Action Detection (TPAMI 2025 · similarity 0.89)
- Weakly-Supervised Temporal Action Detection for Fine-Grained Videos with Hierarchical Atomic Actions (ECCV 2022 · similarity 0.89)
- SCT: Set Constrained Temporal Transformer for Set Supervised Action Segmentation (CVPR 2020 · similarity 0.88)
- Temporal Action Detection With Multi-Level Supervision (ICCV 2021 · similarity 0.88)
- WOAD: Weakly Supervised Online Action Detection in Untrimmed Videos (CVPR 2021 · 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.
