Research map / Video / Action recognition
Revisiting Foreground and Background Separation in Weakly-supervised Temporal Action Localization: A Clustering-based Approach
Qinying Liu, Zilei Wang, Shenghai Rong, Junjie Li, Yixin Zhang
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 snippet clustering component (SCC) joint training (ablation) | WTAL baseline · THUMOS14 | AVG mAP: 42.1 -> 43.2 | Improved |
| add cluster classification component (CCC) (ablation) | baseline + SCC · THUMOS14 | AVG mAP: 43.2 -> 43.9 | Improved |
| add clustering-assisted testing (ablation) | CASE · THUMOS14 | AVG mAP: 43.9 -> 46.2 | Improved |
| use optimal transport marginal constraint to generate pseudo-labels (ablation) | SCC · THUMOS14 | AVG mAP: 40.8 -> 44.7 (hard); 41.2 -> 46.2 (soft) | Improved |
| use soft instead of hard OT labels (ablation) | SCC · THUMOS14 | AVG mAP: 44.7 -> 46.2 | Improved |
| use K=2 clusters instead of many clusters (ablation) | CASE · THUMOS14 | AVG mAP | Hurt |
| increase K beyond 16 (ablation) | CASE · THUMOS14 | AVG mAP: stable within +/-0.2% | No clear effect |
| add Gaussian prior distribution (KL term) to SCC labeling (ablation) | SCC · THUMOS14 | AVG mAP: 45.4 -> 46.2 | Improved |
| estimate CCC marginal from snippet foreground labels instead of equipartition (ablation) | CCC · THUMOS14 | AVG mAP: 44.0 -> 46.2 | Improved |
| fuse attention and cluster-based foreground probabilities at test time (ablation) | CASE · THUMOS14 | AVG mAP: 43.9 / 44.1 -> 46.2 | Improved |
| add CASE to other WTAL baselines | CO2-Net, AICL, Zhou et al. · THUMOS14 | AVG mAP: 44.6 -> 46.6; 46.4 -> 48.0; 48.3 -> 49.2 | 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.
- ASM-Loc — ASM-Loc: Action-Aware Segment Modeling for Weakly-Supervised Temporal Action Localization (CVPR 2022) · 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
- Weakly-Supervised Temporal Action Localization by Inferring Salient Snippet-Feature (AAAI 2024 · similarity 0.94)
- ACSNet: Action-Context Separation Network for Weakly Supervised Temporal Action Localization (AAAI 2021 · similarity 0.93)
- Weakly Supervised Temporal Action Localization via Representative Snippet Knowledge Propagation (CVPR 2022 · similarity 0.92)
- Learning Temporal Co-Attention Models for Unsupervised Video Action Localization (CVPR 2020 · similarity 0.92)
- CoLA: Weakly-Supervised Temporal Action Localization With Snippet Contrastive Learning (CVPR 2021 · similarity 0.92)
- PivoTAL: Prior-Driven Supervision for Weakly-Supervised Temporal Action Localization (CVPR 2023 · similarity 0.91)
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.
