Research map / Video / Action recognition
ReAct: Temporal Action Detection with Relational Queries
Dingfeng Shi, Yujie Zhong, Qiong Cao, Jing Zhang, Lin Ma, Jia Li, Dacheng Tao
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 Relational Attention with IoU Decay (RAID) (ablation) | ReAct · THUMOS14 | average mAP: 51.3 -> 55.0 | Improved |
| add Action Classification Enhancement (ACE) losses (ablation) | ReAct · THUMOS14 | average mAP: 52.1 -> 55.0 | Improved |
| add Segment Quality prediction (ablation) | ReAct · THUMOS14 | average mAP: 52.2 -> 55.0 | Improved |
| add all three components (RAID, ACE, SQ) to the base (ablation) | ReAct base · THUMOS14 | average mAP: 47.7 -> 55.0 | Improved |
| add ACE-enc contrastive loss with positives S1,S2 and negatives N1,N2 (ablation) | ReAct · THUMOS14 | average mAP: 53.1 -> 54.3 | Improved |
| use N2 (containing, low IoU) negatives instead of N1 (different class) only (ablation) | ACE-enc · THUMOS14 | average mAP: 53.4 -> 53.9 | Improved |
| apply ACE-enc after the Transformer encoder instead of before (ablation) | ACE-enc · THUMOS14 | average mAP: 54.1 -> 51.2 | Hurt |
| combine query focal loss and ground-truth focal loss in ACE-dec (ablation) | ACE-dec · THUMOS14 | average mAP: 52.1 -> 53.4 | Improved |
| add same-category positives from other videos (S2) to S1 (ablation) | ACE-enc · THUMOS14 | average mAP: 54.1 -> 54.3 | Improved |
| vary gamma similarity threshold in RAID (ablation) | RAID · THUMOS14 | average mAP | Mixed |
Closest papers by meaning
- Dual DETRs for Multi-Label Temporal Action Detection (CVPR 2024 · similarity 0.90)
- TE-TAD: Towards Full End-to-End Temporal Action Detection via Time-Aligned Coordinate Expression (CVPR 2024 · similarity 0.89)
- Scaling Action Detection: AdaTAD++ with Transformer-Enhanced Temporal-Spatial Adaptation (ICCV 2025 · similarity 0.88)
- DiGIT: Multi-Dilated Gated Encoder and Central-Adjacent Region Integrated Decoder for Temporal Action Detection Transformer (CVPR 2025 · similarity 0.87)
- End-to-End Temporal Action Detection with 1B Parameters Across 1000 Frames (CVPR 2024 · similarity 0.86)
- Prediction-Feedback DETR for Temporal Action Detection (AAAI 2025 · 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.
