Research map / Video / Action recognition
Sports Video Analysis on Large-Scale Data
Dekun Wu, He Zhao, Xingce Bao, Richard P. Wildes
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 |
|---|---|---|---|
| use a transformer decoder model instead of an LSTM | video captioning model with S3D · NSVA test (captioning) | CIDEr: 0.500 -> 0.572 | Improved |
| replace S3D with TimeSformer video features (ablation) | UniVL · NSVA test (captioning) | CIDEr: 0.717 -> 0.956 | Improved |
| add ball, basket, player-with-ball and position-aware features to S3D (ablation) | UniVL (S3D) · NSVA test (captioning) | CIDEr: 0.717 -> 0.986 | Improved |
| add all four finer features (BAL, BAS, PB, PA) to TimeSformer (ablation) | Ours (T) · NSVA test (captioning) | CIDEr: 0.956 -> 1.139 | Improved |
| add position-aware (court line) feature to TimeSformer (ablation) | Ours (T) · NSVA test (captioning) | CIDEr: 0.956 -> 1.064 | Improved |
| add player-with-ball feature to TimeSformer (ablation) | Ours (T) · NSVA test (captioning) | CIDEr: 0.956 -> 1.055 | Improved |
| add basket feature to TimeSformer (ablation) | Ours (T) · NSVA test (captioning) | CIDEr: 0.956 -> 1.023 | Improved |
| add ball feature to TimeSformer (ablation) | Ours (T) · NSVA test (captioning) | CIDEr: 0.956 -> 0.931 | Mixed |
| add basket and position-aware features to TimeSformer with player-with-ball (ablation) | TimeSformer · NSVA action recognition (Action-C) | accuracy: 58.13 -> 60.31 | Improved |
| add position-aware feature (ablation) | TimeSformer+PB+BAL+BAS · NSVA action recognition (Action-C) | accuracy: 60.79 -> 61.20 | Improved |
| add position-aware feature (ablation) | TimeSformer+PB+BAL+BAS · NSVA action recognition (Action-E) | accuracy: 42.29 -> 42.34 | No clear effect |
| use TimeSformer instead of S3D features (ablation) | action recognition · NSVA action recognition (Action-F) | accuracy: 48.81 -> 49.66 | Improved |
Closest papers by meaning
- FineSports: A Multi-person Hierarchical Sports Video Dataset for Fine-grained Action Understanding (CVPR 2024 · similarity 0.88)
- BASKET: A Large-Scale Video Dataset for Fine-Grained Skill Estimation (CVPR 2025 · similarity 0.88)
- Towards Universal Soccer Video Understanding (CVPR 2025 · similarity 0.88)
- ViSTec: Video Modeling for Sports Technique Recognition and Tactical Analysis (AAAI 2024 · similarity 0.87)
- SportR: A Benchmark for Multimodal Large Language Model Reasoning in Sports (ICLR 2026 · similarity 0.86)
- Precise Event Spotting in Sports Videos: Solving Long-Range Dependency and Class Imbalance (CVPR 2025 · 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.
