AI-POWERED PERSONAL FITNESS COACH USING DEEP LEARNING

Abstract

With the increasing awareness of health and fitness, there is a growing demand for intelligent systems that can guide users in performing exercises correctly and safely. This project presents an AI-powered personal fitness coach that utilizes deep learning techniques to recognize exercise activities and provide personalized workout recommendations. Due to the lack of comprehensive datasets containing user-specific parameters such as age, weight, and fitness goals, the system focuses on activity recognition using a Convolutional Neural Network (CNN) model based on the YOLO architecture. The proposed system is trained to recognize 36 different exercise activities from images. When a user uploads an image, the trained model detects the activity being performed and provides appropriate exercise recommendations based on user attributes such as age. Additionally, the system offers guidance on correct posture and execution of the detected exercise, helping users avoid injuries and improve performance. The system includes modules for user login, model training, training visualization, and activity recognition with plan recommendation. Experimental results show that the deep learning model achieves high accuracy in activity recognition, with training graphs indicating increased precision and reduced loss over epochs. The system provides real-time feedback and personalized plans, making it suitable for home-based fitness applications. However, limitations such as dependency on image quality and lack of real-time video processing remain. Overall, this project demonstrates the potential of deep learning in building intelligent fitness coaching systems.