Research map / Video / Action recognition
Uncertainty-Based Spatial-Temporal Attention for Online Action Detection
Hongji Guo, Zhou Ren, Yi Wu, Gang Hua, Qiang Ji
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 uncertainty-based spatial attention (ablation) | TRN · THUMOS-14 (ActivityNet features) | mAP (%): 47.2 -> 48.3 | Improved |
| add uncertainty-based temporal attention (ablation) | TRN · THUMOS-14 (ActivityNet features) | mAP (%): 47.2 -> 50.1 | Improved |
| add joint spatial-temporal attention (ablation) | TRN · THUMOS-14 (ActivityNet features) | mAP (%): 47.2 -> 51.3 | Improved |
| add joint spatial-temporal attention (ablation) | OadTR · THUMOS-14 (Kinetics features) | mAP (%): 65.2 -> 67.5 | Improved |
| add joint spatial-temporal attention (ablation) | LSTR · THUMOS-14 (ActivityNet features) | mAP (%): 65.3 -> 66.0 | Improved |
| add joint spatial-temporal attention (ablation) | LSTR · THUMOS-14 (Kinetics features) | mAP (%): 69.5 -> 69.9 | No clear effect |
| add uncertainty-based STA | TRN · TVSeries (ActivityNet features) | mcAP (%): 83.7 -> 85.2 | Improved |
| add uncertainty-based STA | LSTR · TVSeries (ActivityNet features) | mcAP (%): 88.1 -> 88.3 | No clear effect |
| add uncertainty-based STA | OadTR · HDD | mAP (%): 29.8 -> 30.1 | Improved |
| add uncertainty-based STA | TRN · TVSeries viewpoint generalization | accuracy (%): 65.8 -> 69.5 | Improved |
| add uncertainty-based STA | TRN · TVSeries occlusion generalization | accuracy (%): 85.2 -> 88.6 | Improved |
| add uncertainty-based STA (ablation) | TRN · THUMOS-14 small training data (10%) | mAP | No clear effect |
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.
- OadTR — OadTR: Online Action Detection With Transformers (ICCV 2021) · better on 5 of 5 shared results
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
- Towards Robust Temporal Action Detection: Benchmark and A Strong Baseline (IJCV 2026 · similarity 0.84)
- Uncertainty-Aware Weakly Supervised Action Detection from Untrimmed Videos (ECCV 2020 · similarity 0.82)
- Online Action Detection (ECCV 2016 · similarity 0.82)
- Learning Endogenous Attention for Incremental Object Detection (CVPR 2025 · similarity 0.81)
- Object Priors for Classifying and Localizing Unseen Actions (IJCV 2021 · similarity 0.81)
- ActionSwitch: Class-agnostic Detection of Simultaneous Actions in Streaming Videos (ECCV 2024 · similarity 0.81)
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.
