AI & multi-resolution Temporal Processing for Accurate Counting of Exercises Repetitions

Abstract

Artificial intelligence has its importance in the field of objects' movement tracking, which is extensively applied in the fitness field. Applying machine learning technology in fitness field could introduce an auto judge by counting accurately the repetitions of any exercise. On the other hand, camera resolution depends heavily on number of frames per second (fps) which may affect the AI decision accuracy due to the possibility of losing some frames. In this paper a new algorithm relay on temporal processing for the combination of Computer Vision and deep machine learning techniques are introduced to analyze frames data accurately and report a feedback on the repetitions of exercises.