Reproduce “Extended OpenTT Games Dataset: A table tennis dataset for fine-grained shot type and point outcome” 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
Moamal Fadhil Abdul-Mahdi, Jonas Bruun Hubrechts, Thomas Martini Jørgensen, Emil Hovad
Researchers from DTU Compute expanded the OpenTTGames dataset with detailed, frame-accurate annotations for table tennis, providing specific stroke types, player posture, and rally outcomes. The extended dataset features 1,457 stroke type annotations, 1,432 body lean instances, 1,319 feet position instances, and 282 rally ending classifications, released under an open license to foster advanced tactical and biomechanical analysis.
Key insights
- Providing granular, frame-accurate annotations for aspects like specific stroke types and player posture significantly enhances the potential for machine learning models to achieve a deeper, tactical understanding of table tennis gameplay.
- Openly licensed datasets are crucial for overcoming barriers to innovation and reproducibility in sports analytics, enabling broader research participation and comparative benchmarking.
- The detailed classification of rally-ending events and biomechanical markers (body lean, feet position) at the moment of impact provides new avenues for analyzing strategic intent, performance implications, and skill improvement.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.