Smart Fitness Assistance using AI

Abstract

In computer vision, human body pose estimation is the study that aims at predicting the poses of human body parts and joints. In this project, visual languages are built based on skeleton poses. We present a real-time body tracking pipeline that predicts body posture. Once the joints of the body have been localised to represent a specific part of the human body, we can use them for wider ranges of applications like tracking body movements of some professional athlete or gymnast to understand the various physical techniques and strategies involved to achieve his/her desired goals. It is implemented using MediaPipe, an open-source, cross-platform Machine Learning framework used for building complex and multimodal applied machine learning pipelines. As a result, this system delivers a realtime interface as well as high prediction quality. It might also be utilized to design a smart gym training assistant program that would help individual bodybuilders achieve their intended goals. Following a thorough review of various types of fitness trackers and fitness applications, an evaluation of AI algorithms used in smart fitness scenarios is provided. Finally, a selection of existing gaps and potential future work, as well as extensive lectures on the benefits and potential obstacles of smart fitness, have been identified and offered.