Research map / Video / Action recognition
Learning Action Completeness From Points for Weakly-Supervised Temporal Action Localization
Pilhyeon Lee, Hyeran Byun
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 point-level supervision loss (ablation) | video-level baseline · THUMOS'14 | AVG mAP 0.1:0.7: 28.7 -> 47.3 | Improved |
| add score contrastive loss (ablation) | baseline with point loss · THUMOS'14 | AVG mAP 0.1:0.7: 47.3 -> 52.0 | Improved |
| add feature contrastive loss (ablation) | baseline with point loss · THUMOS'14 | AVG mAP 0.1:0.7: 47.3 -> 49.0 | Improved |
| add feature contrastive loss on top of score loss (ablation) | score-loss model · THUMOS'14 | AVG mAP 0.1:0.7: 52.0 -> 52.8 | Improved |
| score sequences by action contrast instead of inner scores (ablation) | optimal sequence search · THUMOS'14 | AVG mAP 0.1:0.7: 49.0 -> 50.8 | Improved |
| include background instances in score contrast (ablation) | optimal sequence search · THUMOS'14 | AVG mAP 0.1:0.7: 50.8 -> 52.8 | Improved |
| increase greedy search budget from 1 to 25 (ablation) | optimal sequence search · THUMOS'14 | AVG mAP 0.1:0.7: 51.3 -> 52.8 | Improved |
| increase greedy search budget beyond 5 (ablation) | optimal sequence search · THUMOS'14 | AVG mAP 0.1:0.7: 52.5 -> 52.8 | No clear effect |
| use at-least-one background point per section instead of global mining (ablation) | baseline without completeness learning · THUMOS'14 | AVG mAP 0.1:0.7: 43.6 -> 46.6 | Improved |
| add filling between background points (ablation) | pseudo background mining · THUMOS'14 | AVG mAP 0.1:0.7: 46.6 -> 47.3 | Improved |
| use uniformly distributed point labels instead of Gaussian (ablation) | Ours · THUMOS'14 | AVG mAP 0.1:0.7: 52.8 -> 49.3 | Hurt |
| use point-level labels instead of video-level labels | Ours vs Lee et al. · ActivityNet 1.3 | AVG mAP: 23.7 -> 25.1 | 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.
- CoLA — CoLA: Weakly-Supervised Temporal Action Localization With Snippet Contrastive Learning (CVPR 2021) · better on 1 of 1 shared result
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
- Completeness Modeling and Context Separation for Weakly Supervised Temporal Action Localization (CVPR 2019 · similarity 0.94)
- Improving Weakly Supervised Temporal Action Localization by Bridging Train-Test Gap in Pseudo Labels (CVPR 2023 · similarity 0.93)
- Bridge the Gap: From Weak to Full Supervision for Temporal Action Localization with PseudoFormer (CVPR 2025 · similarity 0.91)
- Weakly-Supervised Temporal Action Localization by Inferring Salient Snippet-Feature (AAAI 2024 · similarity 0.90)
- Similar Modality Enhancement and Action Consistency Learning for Weakly Supervised Temporal Action Localization (AAAI 2025 · similarity 0.89)
- 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.
