The dataset contains short video clips of four shoulder exercises.
The videos are labeled as either correct or incorrect.