Reproduce “SportsPose -- A Dynamic 3D sports pose dataset” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Christian Keilstrup Ingwersen, Christian Mikkelstrup, Janus Nørtoft Jensen, Morten Rieger Hannemose, Anders Bjorholm Dahl
Researchers from the Technical University of Denmark and TrackMan A/S developed SportsPose, a large-scale, markerless 3D human pose dataset featuring over 176,000 poses of highly dynamic sports movements from 24 subjects. The dataset's accuracy was quantitatively validated against a commercial marker-based system, showing a mean error of 34.5 mm.
Key insights
- Highly dynamic sports movements can be effectively captured and validated in a large-scale, markerless 3D pose dataset using a scalable multi-camera setup.
- The novel "local movement" metric effectively quantifies the dynamic range and variety of poses in extremum joints, demonstrating SportsPose's superior dynamism compared to existing datasets.
- Rigorous in-domain quantitative validation against marker-based systems provides a reliable approach to ensure the accuracy and trustworthiness of markerless 3D pose ground truth data.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.