Research map / Video / Action recognition
HAT: History-Augmented Anchor Transformer for Online Temporal Action Localization
Sakib Reza, Yuexi Zhang, Mohsen Moghaddam, Octavia Camps
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 the future-supervised history module (long-term history) to the anchor transformer (ablation) | HAT · EGTEA split-1 | Avg. mAP (%): 18.9 -> 20.8 | Improved |
| add action anticipation head to the history module (ablation) | HAT · EGTEA split-1 | Avg. mAP (%): 19.8 -> 20.8 | Improved |
| add future-driven history refinement (ablation) | HAT · EGTEA split-1 | Avg. mAP (%): 20.0 -> 20.8 | Improved |
| use FSHM instead of the LSTR two-stage history compression (ablation) | HAT · EGTEA split-1 | Avg. mAP (%): 19.5 -> 20.8 | Improved |
| use FSHM instead of GateHUB gated history module (ablation) | HAT · EGTEA split-1 | Avg. mAP (%): 19.9 -> 20.8 | Improved |
| use adaptive focal loss instead of cross-entropy (ablation) | HAT · EGTEA split-1 | Avg. mAP (%): 19.4 -> 20.8 | Improved |
| use adaptive focal loss instead of background-suppression focal loss (ablation) | HAT · EGTEA split-1 | Avg. mAP (%): 20.2 -> 20.8 | Improved |
| HAT vs OAT on procedural egocentric datasets | HAT · EGTEA (PREGO) | Avg. mAP (%): 19.6 -> 21.5 | Improved |
| HAT vs OAT on non-procedural THUMOS'14 | HAT · THUMOS'14 | Avg. mAP (%): 44.6 -> 44.8 | No clear effect |
| HAT vs OAT at tIoU 0.3 on THUMOS'14 | HAT · THUMOS'14 | mAP@0.3 (%): 63.0 -> 62.0 | Hurt |
| HAT vs reconstructed OAT on MUSES | HAT · MUSES | Avg. mAP (%): 9.8 -> 10.8 | Improved |
Closest papers by meaning
- A Sliding Window Scheme for Online Temporal Action Localization (ECCV 2022 · similarity 0.90)
- Learning Salient Boundary Feature for Anchor-free Temporal Action Localization (CVPR 2021 · similarity 0.89)
- Weakly-Supervised Temporal Action Localization by Inferring Salient Snippet-Feature (AAAI 2024 · similarity 0.86)
- HR-Pro: Point-Supervised Temporal Action Localization via Hierarchical Reliability Propagation (AAAI 2024 · similarity 0.86)
- HyrCap: Hybrid Rank-Calibration of Action Proposals for Temporal Event Understanding (NeurIPS 2026 · similarity 0.85)
- Realigning Confidence with Temporal Saliency Information for Point-Level Weakly-Supervised Temporal Action Localization (CVPR 2024 · similarity 0.85)
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.
