Research map / Video / Action recognition
Actions as Moving Points
Yixuan Li, Zixu Wang, Limin Wang, Gangshan Wu
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 Movement Branch (center movement) (ablation) | MOC · UCF101-24 (RGB only, K=5) | video mAP@0.5 (%): 37.77 -> 49.55 | Improved |
| add Movement Branch (center movement) (ablation) | MOC · UCF101-24 (RGB only, K=5) | frame mAP@0.5 (%): 68.22 -> 71.63 | Improved |
| add Bbox Align (box estimated at moved center) (ablation) | MOC · UCF101-24 (RGB only, K=5) | video mAP@0.5 (%): 48.82 -> 49.55 | Improved |
| use Center Movement (3D conv) instead of optical-flow guided movement (ablation) | MOC · UCF101-24 (RGB only, K=5) | video mAP@0.5 (%): 42.28 -> 49.55 | Improved |
| use Center Movement instead of cost volume movement (ablation) | MOC · UCF101-24 (RGB only, K=5) | video mAP@0.5 (%): 43.67 -> 49.55 | Improved |
| regress movement relative to key frame instead of accumulating consecutive movement (ablation) | MOC · UCF101-24 (RGB only, K=5) | video mAP@0.5 (%): 46.19 -> 49.55 | Improved |
| increase tubelet length K from 1 to 5 (ablation) | MOC · UCF101-24 (RGB only) | video mAP@0.5 (%): 31.50 -> 49.55 | Improved |
| increase tubelet length K from 5 to 7 (ablation) | MOC · UCF101-24 (RGB only) | video mAP@0.5 (%): 49.55 -> 51.02 | Improved |
| increase tubelet length K from 7 to 9 (ablation) | MOC · UCF101-24 (RGB only) | video mAP@0.5 (%): 51.02 -> 50.16 | Hurt |
| feed stacked temporal features to Box Branch (ablation) | MOC · UCF101-24 (RGB only, K=5) | video mAP@0.5 (%): 49.55 -> 49.14 | Hurt |
| COCO pretrain instead of ImageNet pretrain | MOC · JHMDB | frame mAP@0.5 (%): 68.0 -> 70.8 | Improved |
| COCO pretrain instead of ImageNet pretrain | MOC · UCF101-24 | frame mAP@0.5 (%): 76.9 -> 78.0 | Improved |
Closest papers by meaning
- Action Tubelet Detector for Spatio-Temporal Action Localization (ICCV 2017 · similarity 0.91)
- TubeR: Tubelet Transformer for Video Action Detection (CVPR 2022 · similarity 0.89)
- End-to-End Spatio-Temporal Action Localisation with Video Transformers (CVPR 2024 · similarity 0.86)
- CFAD: Coarse-to-Fine Action Detector for Spatiotemporal Action Localization (ECCV 2020 · similarity 0.86)
- Finding Action Tubes with a Sparse-to-Dense Framework (AAAI 2020 · similarity 0.86)
- Recurrent Tubelet Proposal and Recognition Networks for Action Detection (ECCV 2018 · similarity 0.86)
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.
