My research work focuses on classifying human exercises movements from videos into correct/incorrect classes. To achieve this, we leverage human pose estimation libraries in conjunction with multivariate time series methods. We have successfully validated our methodology on real-world datasets, consisting of video recordings of two widely practiced strength and conditioning exercises: the Military Press and Rowing.
This work holds particular promise in providing accurate and timely feedback to physiotherapists, coaches, and rehabilitation patients following the execution of physical exercises. By bridging the gap between video analysis and exercise assessment, our research aims to significantly benefit those involved in fitness training and rehabilitation.