Researchers have introduced CalTennis, a large-scale video dataset designed for evaluating monocular-to-3D human pose estimation. The dataset features over 11 million frames of tennis play from 40 players, captured using multiple synchronized cameras. This benchmark is significantly larger than existing datasets and enables label-free evaluation of pose estimation algorithms, highlighting current model limitations in depth and foot contact estimation while proposing new metrics for performance analysis. AI
IMPACT Provides a large-scale benchmark for advancing monocular-to-3D pose estimation, potentially improving applications in sports analytics and human-computer interaction.
RANK_REASON The cluster describes a new dataset and benchmark published on arXiv for computer vision research.
- arXiv
- CalTennis
- Monocular-to-3D Pose Estimation
- alphaXiv
- Caltech Tennis Dataset
- DagsHub
- Hugging Face
- motion capture
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
