Research map / Video / Action recognition
OadTR: Online Action Detection With Transformers
Xiang Wang, Shiwei Zhang, Zhiwu Qing, Yuanjie Shao, Zhengrong Zuo, Changxin Gao, Nong Sang
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 task token to encoder-only baseline (ablation) | OadTR encoder · THUMOS14 | mAP: 55.8 -> 56.9 | Improved |
| add task token to encoder-only baseline (ablation) | OadTR encoder · HDD | mAP: 28.7 -> 29.0 | Improved |
| add task token to encoder-only baseline (ablation) | OadTR encoder · TVSeries | mcAP: 84.8 -> 85.0 | Improved |
| add future-prediction decoder to encoder-only baseline (ablation) | OadTR encoder · THUMOS14 | mAP: 55.8 -> 56.7 | Improved |
| add future-prediction decoder to encoder-only baseline (ablation) | OadTR encoder · HDD | mAP: 28.7 -> 29.2 | Improved |
| add both task token and decoder to encoder-only baseline (ablation) | OadTR encoder · THUMOS14 | mAP: 55.8 -> 58.3 | Improved |
| use learned position encoding instead of none (ablation) | OadTR · HDD | mAP: 28.8 -> 29.8 | Improved |
| use learned instead of fixed position encoding (ablation) | OadTR · HDD | mAP: 29.3 -> 29.8 | Improved |
| use avg-pool instead of max-pool to aggregate decoder outputs (ablation) | OadTR · THUMOS14 | mAP: 57.9 -> 58.3 | Improved |
| add encoder feature encoding (ablation) | OadTR · THUMOS14 | mAP: 53.5 -> 58.3 | Improved |
| replace Sparse Transformer with vanilla Transformer (ablation) | OadTR · THUMOS14 | mAP: 58.1 -> 58.3 | No clear effect |
| use 8 decoder steps instead of 2 (ablation) | OadTR · THUMOS14 | mAP: 57.4 -> 58.3 | Improved |
Later papers that compared with this method
5 later accepted papers report OadTR as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- Backtrace Mamba: Reviving Critical Temporal Contexts via Hierarchical Memory Compression for Online Action Detection (AAAI 2026 · better on 1 of 1 shared result)
- Colar: Effective and Efficient Online Action Detection by Consulting Exemplars (CVPR 2022 · better on 7 of 7 shared results)
- Uncertainty-Based Spatial-Temporal Attention for Online Action Detection (ECCV 2022 · better on 5 of 5 shared results)
- Real-Time Online Video Detection with Temporal Smoothing Transformers (ECCV 2022 · better on 2 of 2 shared results)
- GateHUB: Gated History Unit With Background Suppression for Online Action Detection (CVPR 2022 · better on 1 of 1 shared result)
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
- MiniROAD: Minimal RNN Framework for Online Action Detection (ICCV 2023 · similarity 0.88)
- Long Short-Term Transformer for Online Action Detection (NeurIPS 2021 · similarity 0.88)
- Does Video-Text Pretraining Help Open-Vocabulary Online Action Detection? (NeurIPS 2024 · similarity 0.88)
- Context-Enhanced Memory-Refined Transformer for Online Action Detection (CVPR 2025 · similarity 0.88)
- Learning to Discriminate Information for Online Action Detection (CVPR 2020 · similarity 0.88)
- Dual DETRs for Multi-Label Temporal Action Detection (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.
