Research map / Video / Action recognition
ACGNet: Action Complement Graph 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 |
|---|---|---|---|
| Add ACGNet to BaS-Net | BaS-Net · THUMOS'14 test | Average mAP (%): 34.9 -> 37.0 (+2.1) | Helped |
| Add ACGNet to DGAM | DGAM · THUMOS'14 test | Average mAP (%): 36.3 -> 38.2 (+1.9) | Helped |
| Add ACGNet to BaM | BaM · THUMOS'14 test | Average mAP (%): 41.2 -> 42.5 (+1.3) | Helped |
| Add ACGNet to BaS-Net | BaS-Net · ActivityNet1.2 validation | Average mAP (%): 22.4 -> 25.1 (+2.7) | Helped |
| Add ACGNet to DGAM | DGAM · ActivityNet1.2 validation | Average mAP (%): 24.0 -> 24.9 (+0.9) | Helped |
| Add ACGNet to BaM | BaM · ActivityNet1.2 validation | Average mAP (%): 25.6 -> 26.1 (+0.5) | Helped |
| Combine similarity and temporal diffusion graphs instead of using similarity graph only (ablation) | BaS-Net + ACGNet · THUMOS'14 test | mAP (%) at IoU=0.5, K=50: 28.0 -> 29.8 | Helped |
| Use a dense graph without sparsification (ablation) | BaS-Net + ACGNet · THUMOS'14 test | mAP (%) at IoU=0.5: 29.8 -> 25.3 | Hurt |
| Use graph-average features alone instead of original features (ablation) | BaS-Net + ACGNet · THUMOS'14 test | mAP (%) at IoU=0.5: 26.8 -> 19.7 | Hurt |
| Add graph-average features to original features using concatenation (ablation) | BaS-Net + ACGNet · THUMOS'14 test | mAP (%) at IoU=0.5: 26.8 -> 29.1 | Helped |
| Add graph-convolution features to original and average features (ablation) | BaS-Net + ACGNet · THUMOS'14 test | mAP (%) at IoU=0.5: 28.2 -> 29.8 | Helped |
| Remove EPM loss from the full feature configuration (ablation) | BaS-Net + ACGNet · THUMOS'14 test | mAP (%) at IoU=0.5: 29.8 -> 28.2 | Hurt |
Closest papers by meaning
- ACSNet: Action-Context Separation Network for Weakly Supervised Temporal Action Localization (AAAI 2021 · similarity 0.91)
- Temporal Action Localization in Untrimmed Videos via Multi-Stage CNNs (CVPR 2016 · similarity 0.90)
- G-TAD: Sub-Graph Localization for Temporal Action Detection (CVPR 2020 · similarity 0.90)
- ASM-Loc: Action-Aware Segment Modeling for Weakly-Supervised Temporal Action Localization (CVPR 2022 · similarity 0.90)
- Weakly-Supervised Temporal Action Localization by Inferring Salient Snippet-Feature (AAAI 2024 · similarity 0.90)
- Bridge the Gap: From Weak to Full Supervision for Temporal Action Localization with PseudoFormer (CVPR 2025 · 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.