Research map / Video / Action recognition
Background-Click Supervision for Temporal Action Localization
Le Yang, Junwei Han, Tao Zhao, Tianwei Lin, Dingwen Zhang, Jianxin Chen
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 background-click supervision (ablation) | baseline · THUMOS14 | mAP@tIoU0.5: 28.6 -> 35.0 | Improved |
| add action-click supervision (ablation) | baseline · THUMOS14 | mAP@tIoU0.5: 28.6 -> 29.1 | No clear effect |
| add background clicks on top of action clicks (ablation) | baseline + action click · THUMOS14 | mAP@tIoU0.5: 29.1 -> 36.8 | Improved |
| add score separation module (ablation) | baseline + background click · THUMOS14 | mAP@tIoU0.5: 35.0 -> 35.6 | Improved |
| add affinity module (ablation) | baseline + background click · THUMOS14 | mAP@tIoU0.5: 35.0 -> 35.8 | Improved |
| add both score separation and affinity modules (ablation) | baseline + background click · THUMOS14 | mAP@tIoU0.5: 35.0 -> 36.3 | Improved |
| add weight supervision to CAS supervision (ablation) | baseline + CAS supervision · THUMOS14 | mAP@tIoU0.5: 35.0 -> 35.2 | No clear effect |
| increase neighboring frame number from 3 to 9 (ablation) | BackTAL · THUMOS14 | mAP@tIoU0.5: 36.3 -> 35.8 | Hurt |
| reduce embedding dimension from 32 to 8 (ablation) | BackTAL · THUMOS14 | mAP@tIoU0.5: 36.3 -> 35.7 | Hurt |
| increase tau_same from 0.5 to 0.7 (ablation) | BackTAL · THUMOS14 | mAP@tIoU0.5: 36.3 -> 35.6 | Hurt |
| decrease lambda from 1.0 to 0.8 (ablation) | BackTAL · THUMOS14 | mAP@tIoU0.5: 36.3 -> 35.8 | Hurt |
| use action-click annotations with BackTAL modules instead of SF-Net | SF-Net · THUMOS14 | mAP@tIoU0.5: 30.5 -> 32.4 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ACSNet — ACSNet: Action-Context Separation Network for Weakly Supervised Temporal Action Localization (AAAI 2021) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Temporal Action Localization in the Deep Learning Era: A Survey (TPAMI 2023 · similarity 0.93)
- PivoTAL: Prior-Driven Supervision for Weakly-Supervised Temporal Action Localization (CVPR 2023 · similarity 0.93)
- SF-Net: Single-Frame Supervision for Temporal Action Localization (ECCV 2020 · similarity 0.92)
- Weakly-Supervised Temporal Action Localization by Inferring Salient Snippet-Feature (AAAI 2024 · similarity 0.92)
- Bridge the Gap: From Weak to Full Supervision for Temporal Action Localization with PseudoFormer (CVPR 2025 · similarity 0.91)
- Realigning Confidence with Temporal Saliency Information for Point-Level Weakly-Supervised Temporal Action Localization (CVPR 2024 · similarity 0.90)
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.
