Research map / Video / Action recognition
SportsCap: Monocular 3D Human Motion Capture and Fine-Grained Understanding in Challenging Sports Videos
Xin Chen, Anqi Pang, Wei H. Yang, Yuexin Ma, Lan Xu, Jingyi Yu A
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 prior loss from sports pose embedding space (ablation) | Motion Embedding Module · SMART | PCK-0.3: 70.9 -> 83.6 | Improved |
| use semantic (per sub-motion) rather than general pose space (ablation) | Motion Embedding Module · SMART | PCK-0.3: 83.0 -> 84.8 | Improved |
| use semantic (per sub-motion) rather than general pose space (ablation) | Motion Embedding Module · SMART | PCK-0.5: 92.1 -> 92.4 | Improved |
| add multi-task learning with action parsing module (ablation) | Motion Embedding Module (ResNet152) · SMART | PCK-0.5: 94.8 -> 96.0 | Improved |
| deeper backbone ResNet50 to ResNet101 (ablation) | Motion Embedding Module · SMART | PCK-0.3: 84.8 -> 87.5 | Improved |
| deeper backbone ResNet101 to ResNet152 (ablation) | Motion Embedding Module · SMART | PCK-0.3: 87.5 -> 88.5 | Improved |
| add pose-coefficient P-stream to joint and bone streams with SAMB (ablation) | Action Parsing Module · SMART diving number | Top-1 accuracy: 67.1 -> 82.2 | Improved |
| add bone stream to joint stream (ablation) | Action Parsing Module · SMART diving number | Top-1 accuracy: 65.0 -> 67.1 | Improved |
| use Semantic Attributes Mapping Block instead of black-box (ablation) | Action Parsing Module · SMART diving number | Top-1 accuracy: 78.0 -> 82.2 | Improved |
| add SAMB to speed convergence (ablation) | Action Parsing Module · SMART | epochs to converge: 70 -> 50 -> 10 | Improved |
| use SportsCap instead of ST-GCN | Action parsing · FineGym partial Gym288 | mean accuracy: 11.0 -> 46.9 | Improved |
| use SportsCap instead of C3D-AVG | Action parsing · AQA | Twist No. / Some No.: 97.5 -> 97.9 / 96.9 -> 96.3 | Mixed |
Closest papers by meaning
- MultiSports: A Multi-Person Video Dataset of Spatio-Temporally Localized Sports Actions (ICCV 2021 · similarity 0.89)
- FineSports: A Multi-person Hierarchical Sports Video Dataset for Fine-grained Action Understanding (CVPR 2024 · similarity 0.89)
- ViSTec: Video Modeling for Sports Technique Recognition and Tactical Analysis (AAAI 2024 · similarity 0.87)
- SportsHHI: A Dataset for Human-Human Interaction Detection in Sports Videos (CVPR 2024 · similarity 0.87)
- ActionAtlas: A VideoQA Benchmark for Domain-specialized Action Recognition (NeurIPS 2024 · similarity 0.86)
- Learning Skill-Attributes for Transferable Assessment in Video (NeurIPS 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.
