Research map / Video / Action recognition
ACSNet: Action-Context Separation Network for Weakly Supervised Temporal Action Localization
Read the paperCheck 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 |
|---|---|---|---|
| Separate action from context using the ACSNet action-context branch | WS-TAL with UntrimmedNet features · THUMOS14 test | average mAP@IoU 0.3–0.7: 22.6 -> 25.4 | Helped |
| Separate action from context using the ACSNet action-context branch | WS-TAL with I3D features · THUMOS14 test | average mAP@IoU 0.3–0.7: 29.0 -> 32.0 | Helped |
| Separate action from context using ACSNet | WS-TAL · ActivityNet v1.2 validation | average mAP@IoU 0.5:0.05:0.95: 24.4 -> 26.0 | Helped |
| Separate action from context using ACSNet | WS-TAL · ActivityNet v1.3 validation | average mAP@IoU 0.5:0.05:0.95: 22.2 -> 23.9 | Helped |
| Set α to 0.4 instead of 0.5 in the FB-only baseline (ablation) | FB baseline · THUMOS14 test | average mAP@IoU 0.3–0.7: 17.0 -> 22.4 | Helped |
| Add action-attention proposals (P2) to the FB-only baseline (ablation) | ACSNet localization · THUMOS14 test | average mAP@IoU 0.3–0.7: 22.4 -> 25.6 | Helped |
| Add action-context-offset scoring (S2) to P2 proposal generation (ablation) | ACSNet localization · THUMOS14 test | average mAP@IoU 0.3–0.7: 25.6 -> 29.8 | Helped |
| Use action-context offsets for proposal generation (P3) instead of action-attention proposals (P2) (ablation) | ACSNet localization with S2 scoring · THUMOS14 test | average mAP@IoU 0.3–0.7: 29.8 -> 31.3 | Helped |
| Add action-attention proposals (P2) to P3 proposal generation and S2 scoring (ablation) | ACSNet localization · THUMOS14 test | average mAP@IoU 0.3–0.7: 31.3 -> 32.0 | Helped |
| Add FB-generated proposals (P1) to P2 + P3 + S2 (ablation) | ACSNet localization · THUMOS14 test | average mAP@IoU 0.3–0.7: 32.0 -> 28.3 | Hurt |
| Select action-attention snippets rather than all foreground snippets (ablation) | ACSNet snippet localization · THUMOS14 test | average mAP@IoU 0.3–0.7: 22.4 -> 25.6 | Helped |
Closest papers by meaning
- Weakly Supervised Temporal Action Localization Through Learning Explicit Subspaces for Action and Context (AAAI 2021 · similarity 0.94)
- Completeness Modeling and Context Separation for Weakly Supervised Temporal Action Localization (CVPR 2019 · similarity 0.93)
- Weakly-Supervised Temporal Action Localization by Inferring Salient Snippet-Feature (AAAI 2024 · similarity 0.91)
- Temporal Action Localization with Cross Layer Task Decoupling and Refinement (AAAI 2025 · similarity 0.90)
- Realigning Confidence with Temporal Saliency Information for Point-Level Weakly-Supervised Temporal Action Localization (CVPR 2024 · similarity 0.90)
- Test-Time Zero-Shot 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.