Research map / Video / Action recognition
Efficient Action Counting with Dynamic Queries
Xiaoxuan Ma, Zishi Li, Qiuyan Shang, Wentao Zhu, Hai Ci, Yu Qiao, Yizhou Wang
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 Dynamic Action Query (initialise decoder queries from selected encoder features) (ablation) | action counting model · RepCountA test (TSN backbone) | MAE: 0.3542 -> 0.2809 | Improved |
| add Dynamic Action Query (ablation) | action counting model · RepCountA test (TSN backbone) | OBO: 0.4172 -> 0.4570 | Improved |
| add Inter-query Contrastive Learning (ablation) | action counting model · RepCountA test (TSN backbone) | MAE: 0.4035 -> 0.2809 | Improved |
| add Inter-query Contrastive Learning (ablation) | action counting model · RepCountA test (TSN backbone) | OBO: 0.4040 -> 0.4570 | Improved |
| add DAQ to short-period actions (ablation) | action counting model · RepCountA test (TSN backbone) | OBO (short): 0.2093 -> 0.1628 | Hurt |
| switch backbone from TSN to ViT | action counting model · RepCountA test | OBO: 0.4570 -> 0.5430 | Improved |
| switch backbone from I3D to ViT | action counting model · RepCountA test | MAE: 0.3305 -> 0.2622 | Improved |
| use query-based linear-complexity model instead of TransRAC at T=64 | action counting model · RepCountA inference | inference time (s): 2.43 -> 0.15 | Improved |
| increase input length from 64 to 512 frames | action counting model · RepCountA inference | inference time (s): 0.15 -> 0.78 | Improved |
| raise confidence threshold above 0.5 (ablation) | action counting model · RepCountA test | OBO | Hurt |
| use ViT-backed query model instead of TransRAC on unseen UCFRep | action counting model · UCFRep test (trained on RepCountA) | OBO: 0.3143 -> 0.4184 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- TransRAC — TransRAC: Encoding Multi-Scale Temporal Correlation With Transformers for Repetitive Action Counting (CVPR 2022) · better on 7 of 7 shared results
- ActionFormer — ActionFormer: Localizing Moments of Actions with Transformers (ECCV 2022) · better on 2 of 2 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
- Context-Aware and Scale-Insensitive Temporal Repetition Counting (CVPR 2020 · similarity 0.91)
- TransRAC: Encoding Multi-Scale Temporal Correlation With Transformers for Repetitive Action Counting (CVPR 2022 · similarity 0.91)
- CountLLM: Towards Generalizable Repetitive Action Counting via Large Language Model (CVPR 2025 · similarity 0.90)
- Counting Out Time: Class Agnostic Video Repetition Counting in the Wild (CVPR 2020 · similarity 0.89)
- Unsupervised Action Segmentation via Fast Learning of Semantically Consistent Actoms (AAAI 2024 · similarity 0.86)
- FACT: Frame-Action Cross-Attention Temporal Modeling for Efficient Action Segmentation (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.
