Research map / Video / Action recognition
CFAD: Coarse-to-Fine Action Detector for Spatiotemporal Action Localization
Yuxi Li, Weiyao Lin, John See, Ning Xu Shugong Xu, Ke Yan, Cong Yang
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 location refinement module (sigma=0.8) (ablation) | CFAD coarse tubes · UCF101-24, k=2 | v-mAP (%): 46.0 -> 60.3 | Improved |
| add location refinement module (sigma=0.8) (ablation) | CFAD coarse tubes · UCF101-24, k=4 | v-mAP (%): 51.6 -> 62.7 | Improved |
| add location refinement module (sigma=0.8) (ablation) | CFAD coarse tubes · JHMDB-21, k=2 | v-mAP (%): 80.9 -> 83.7 | Improved |
| enlarge refinement search area sigma from 0.4 to 0.8 (ablation) | CFAD · UCF101-24, k=4 | v-mAP (%): 58.8 -> 62.7 | Improved |
| enlarge search area beyond sigma=0.8 (ablation) | CFAD · UCF101-24 | v-mAP (%): less than +0.2% | No clear effect |
| raise polynomial order from 4 to 5 (ablation) | CFAD · UCF101-24 | v-mAP (%): 62.7 -> 61.6 | Hurt |
| raise polynomial order from 2 to 4 (ablation) | CFAD · UCF101-24 | v-mAP (%): 60.3 -> 62.7 | Improved |
| add temporal residual (Diff Block) branch (ablation) | CFAD coarse module · UCF101-24 | wrong classification errors: reduced by 42% | Improved |
| add temporal residual (Diff Block) branch (ablation) | CFAD coarse module · JHMDB-21 | wrong classification errors: reduced by 14% | Improved |
| add refinement (ablation) | CFAD · UCF101-24 | v-MABO: at most +4.7% | Improved |
| use dynamic key timestamp selection instead of random or uniform (ablation) | CFAD · UCF101-24 | v-mAP at similar time cost | Improved |
| add optical flow input to RGB | CFAD · UCF101-24 delta=0.5 | v-mAP (%): 62.7 -> 64.6 | Improved |
Later papers that compared with this method
1 later accepted paper reports CFAD as a baseline in their results tables; 0 of them report a better number on most of the results they share with it.
- TubeR: Tubelet Transformer for Video Action Detection (CVPR 2022 · better on 0 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
- Action Tubelet Detector for Spatio-Temporal Action Localization (ICCV 2017 · similarity 0.91)
- End-to-End Spatio-Temporal Action Localisation with Video Transformers (CVPR 2024 · similarity 0.89)
- Online Real-Time Multiple Spatiotemporal Action Localisation and Prediction (ICCV 2017 · similarity 0.89)
- Finding Action Tubes with a Sparse-to-Dense Framework (AAAI 2020 · similarity 0.89)
- Pointly-Supervised Action Localization (IJCV 2018 · similarity 0.88)
- Improving Action Localization by Progressive Cross-Stream Cooperation (CVPR 2019 · similarity 0.88)
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.
