PosePerfectAI: AI-Based Yoga Pose Detection & Personal Trainer

PosePerfectAI is an AI-based mobile application developed to assist users in analyzing yoga poses using computer vision. The system combines a React Native mobile application with a Python-based AI engine through a FastAPI REST API. OpenCV is used for image processing, while MediaPipe is used to detect human body landmarks for pose analysis. The current implementation provides a complete captured-frame analysis pipeline from the mobile camera to the Python AI engine and back to the mobile application. The system currently supports TREE, WARRIOR, and CHAIR pose analysis and provides information such as pose score, feedback, status, and hold time. The developed system shows how computer vision can be combined with a mobile application to provide posture-related information during yoga practice. The current results demonstrate the working of the implemented pose-analysis pipeline for the tested poses. However, further testing with more users and test cases is required before making broader performance claims. The next major development stage is continuous camera-based pose tracking, which can allow the score and feedback to be updated while the user is performing the pose. With further development and systematic testing, PosePerfectAI can be extended to support more yoga poses and provide improved live posture feedback.

[1] S.-C. Yeh and C.-K. Yang, “Yoga pose recognition and motion analysis for a home-based fitness monitoring and health management system,” Signal, Image and Video Processing, vol. 19, article 841, 2025, doi: 10.1007/s11760-025-04436-6. [2] D. Santhadevi, S. Yerigera, S. Aaditya, and B. R. K. Reddy, “Yoga pose detection and correction using MediaPipe landmarks and a deep multi-layer perceptron classifier,” in Proceedings of the 1st Engineering Data Analytics and Management Conference (EAMCON 2025), Advances in Engineering Research, pp. 332–351, 2025, doi: 10.2991/978-94-6463-978-0_30. [3] R. Bansal, R. Sharma, P. Jain, R. Arora, S. Pal, and Vishal, “DeepYoga: Enhancing Practice with a Real-Time Yoga Pose Recognition System,” Engineering, Technology & Applied Science Research, vol. 14, no. 6, pp. 17704–17710, 2024, doi: 10.48084/etasr.8643. [4] I. Chaudhary, N. T. Singh, M. Chaudhary, and K. Yadav, “Real-Time Yoga Pose Detection Using OpenCV and MediaPipe,” in 2023 4th International Conference for Emerging Technology (INCET), Belgaum, India, pp. 1–5, 2023, doi: 10.1109/INCET57972.2023.10170485.