AI FitCoach - Intelligent Virtual Coach for Safe and Effective Home Workouts

Motion analysis, posture estimation, and human detection are important applications in sports analytics, healthcare, and surveillance. This paper proposes a hybrid system which incorporates MediaPipe Pose for the precise prediction of skeletons, YOLOv8 for accurate human detection, and a rule-based pose correction mechanism for better detection reliability. Long Short-Term Memory (LSTM) networks for motion tracking and activity classification further process the corrected pose data. The proposed architecture offers a real-time and scalable solution in various scenarios to issues posed by accuracy in pose and the prediction of motion. On the basis of experimental results, this system is reported as a major breakthrough in the area of intelligent video analytics due to its detection capability, efficient pose correction, and dependable motion tracking.