Virtual Booth

This project develops a machine learning-based system to analysis cycling form positions on indoor bikes, providing real-time feedback for correction and improvement. The system uses computer vision technology and 3D joints visualization, track joints angle, detecting incorrect movements and contributing to performance enhancement and injury prevention. Trained on a large our dataset, the model achieves 100% accuracy on a testing set of 200-300 cycling performance images. Future developments include integrating high-accuracy pose detection models and expanding the application.