Use of Computer Vision Methods for Guidance in Physical Exercises

A couple of years ago, we witnessed a severe crisis related to the coronavirus. This crisis significantly affected people's exercise habits and activities, whether outdoors or in fitness studios. Due to the inability to visit such places for exercise, more and more people opted for various forms of home workouts. This posed a challenge regarding the proper and safe execution of these exercises, especially for beginners. In this work, we analyze how computer vision methods can assist in recognizing human posture and key body points and how this information can guide and support individuals in performing exercises correctly. Finally, we present our implementation of such an application, which, with the help of pose estimation, provides users with assistance in executing their workouts properly. We named the application Sweatpose.