Reproduce “BlurBall: Joint Ball and Motion Blur Estimation for Table Tennis Ball Tracking” 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

Thomas Gossard, Filip Radovic, Andreas Ziegler, Andreas Zell

The BlurBall framework introduces a novel approach to table tennis ball tracking, jointly estimating ball position and motion blur attributes by leveraging a new blur-aware labeling convention. This method enhances ball detection accuracy, achieving an F1-score of 97.17, and significantly improves trajectory prediction, reducing the Mean Absolute Error from 84.4 to 53.0 pixels by explicitly incorporating velocity information encoded in motion blur.

Key insights

  • Treating motion blur as an informative signal for velocity rather than just noise significantly enhances ball detection and trajectory prediction.
  • A symmetric, blur-aware labeling convention, placing the ground truth at the blur's midpoint, consistently improves detection model performance compared to traditional leading-edge labeling.
  • Explicitly modeling motion blur attributes (length and orientation) provides valuable velocity information, leading to more accurate and robust trajectory predictions even with limited initial frames.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.