Computer Vision for Sports

Description:

Our lab focuses on advancing computer vision applications in sports science, aiming to make athlete performance evaluation more accessible and accurate. We develop tools that use video recordings to estimate key biomechanical variables—such as jump velocity—with high precision, enabling assessments without expensive equipment like force plates. Additionally, we enhance movement quality analysis by automatically detecting athlete execution errors using pose estimation tools such as OpenPose, augmented with expert-labeled datasets. Through this work, we strive to democratize athletic assessments and improve training feedback using scalable, cost-effective computer vision solutions.

Publications:

Roy, S., Roygaga, C., Blanchard, N., & Bharati, A. (2023). A computer vision method for estimating velocity from jumps. In Computer Vision 4 Winter Sports (CV4WS) at Winter Conference for AI. [DOI]

Roygaga, C., *Patil, D., *Boyle, M., Reiser, R., Bharati, A., Blanchard, N. (2022) APE-V: Athlete Performance Evaluation using Video. In Proceedings 2022 IEEE Winter Conference on Applications of Computer Vision (WACV) Workshops. IEEE. [DOI]