FormFit: AI-Based Real-Time Fitness Training and Posture Feedback System

Home workouts have become increasingly popular as they allow individuals to exercise without the need for gym access, personal trainers, or dedicated equipment. However, performing workouts without supervision often results in incorrect posture, reduced efficiency, and a higher risk of injury. To address this concern, the present work introduces a real-time exercise posture evaluation and feedback system that operates directly in a web browser. The system makes use of the MediaPipe BlazePose model to detect human skeletal landmarks from a webcam stream and applies joint-angle calculations to assess posture accuracy across common exercises such as squats, lunges, push-ups, and bicep curls. A rule-based decision mechanism is employed to verify posture correctness and count valid repetitions only when movements meet predefined thresholds. Immediate on-screen feedback assists users in adjusting their form during the workout rather than after completion. Exercise performance data is stored using MongoDB, and a React-based dashboard enables users to review their progress over time. Experimental evaluation shows that the system performs reliably across different body types and indoor conditions. The approach demonstrates that computer-vision-based posture monitoring can provide affordable and effective guidance for home fitness without requiring wearable sensors or external supervision.