Research map / Video / Action recognition
ASM-Loc: Action-Aware Segment Modeling for Weakly-Supervised Temporal Action Localization
Check 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 action-aware segment modeling with proposal refinement | ASM-Loc · THUMOS-14 | average mAP@IoU 0.1:0.7 (%): 40.3 to 45.1 | Helped |
| Use ASM-Loc instead of the strongest reported weakly-supervised baseline | THUMOS-14 weakly-supervised localization · THUMOS-14 | average mAP@IoU 0.1:0.7 (%): 43.6 -> 45.1 | Helped |
| Use ASM-Loc instead of the strongest reported baseline | ActivityNet-v1.3 localization · ActivityNet-v1.3 | average mAP@IoU 0.5:0.05:0.95 (%): 24.4 -> 25.1 | Helped |
| Add background loss to foreground video-level loss (ablation) | Base model · THUMOS-14 | average mAP@IoU 0.1:0.7 (%): 24.3 -> 36.6 | Helped |
| Add action-aware background loss to foreground and background losses (ablation) | Base model · THUMOS-14 | average mAP@IoU 0.1:0.7 (%): 36.6 -> 40.3 | Helped |
| Add dynamic segment sampling (ablation) | ASM-Loc · THUMOS-14 XS actions | averaged mAP (%): 10.6 -> 15.5 (+4.9) | Helped |
| Add dynamic segment sampling (ablation) | ASM-Loc · THUMOS-14 | average mAP@IoU (%): 40.3 -> 41.4 (+1.1) | Helped |
| Apply self-attention globally to all snippets rather than within action proposals (ablation) | ASM-Loc · THUMOS-14 | average mAP@IoU (%): 40.3 -> 40.1 | Hurt |
| Apply self-attention to background snippets only (ablation) | ASM-Loc · THUMOS-14 | average mAP@IoU (%): 40.3 -> 39.6 | Hurt |
| Add proposal-based intra-segment attention (ablation) | ASM-Loc · THUMOS-14 | average mAP@IoU (%): 40.3 -> 41.8 | Helped |
| Add uncertainty estimation to pseudo instance-level supervision (ablation) | ASM-Loc · THUMOS-14 | average mAP@IoU (%): 44.1 -> 45.1 | Helped |
| Increase proposal refinement from 1 to 3 steps (ablation) | ASM-Loc · THUMOS-14 | average mAP@IoU (%): 43.1 -> 45.1 | Helped |
Closest papers by meaning
- PivoTAL: Prior-Driven Supervision for Weakly-Supervised Temporal Action Localization (CVPR 2023 · similarity 0.93)
- Proposal-Based Multiple Instance Learning for Weakly-Supervised Temporal Action Localization (CVPR 2023 · similarity 0.92)
- Weakly-Supervised Temporal Action Localization by Inferring Salient Snippet-Feature (AAAI 2024 · similarity 0.92)
- Learning Temporal Co-Attention Models for Unsupervised Video Action Localization (CVPR 2020 · similarity 0.91)
- Similar Modality Enhancement and Action Consistency Learning for Weakly Supervised Temporal Action Localization (AAAI 2025 · similarity 0.91)
- Bridge the Gap: From Weak to Full Supervision for Temporal Action Localization with PseudoFormer (CVPR 2025 · 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.