Real-Time Exercise Repetition Counting and Form Analysis Using Mediapipe and Computer Vision

Exercise repetition tracking plays a vital role in fitness monitoring and progress evaluation, yet manual counting can be error-prone and inefficient. This research introduces a computer vision-based system that automates the tracking of exercise repetitions using MediaPipe, a tool known for its robust real-time pose estimation capabilities. The system is designed to accurately count repetitions for exercises such as push-ups, squats, and lunges by analyzing body movements. A dataset comprising over 5,000 images of individuals performing these exercises was used to train a deep learning model based on a convolutional neural network (CNN) architecture. To enhance accuracy, transfer learning techniques were employed, allowing the model to adapt to new physical activities. After fine-tuning the model's parameters, the system achieved an overall accuracy rate of 93 % on a test set of 200 images, with specific accuracy rates of 93% for push-ups, 95% for squats, and 98% for lunges. The system also proved resilient to variations in lighting and environmental conditions, demonstrating its reliability for real-time workout monitoring. Further research is required to explore the potential of this system in broader fitness and public health applications.