REAL-TIME YOGA POSE DETECTION AND CORRECTION

Abstract

With the increasing popularity of yoga and its numerous health benefits, it is crucial to ensure that practitioners are able to perform the poses correctly to avoid injury and maximize the benefits. However, traditional methods of learning and practicing yoga often lack real time feedback and guidance.This project addresses the need for an effective and user-friendly solution to enhance the practice of yoga and aims to develop a real-time yoga pose detection system that can accurately analyze and provide feedback on the user’s pose, helping them improve their form and achieve better results. The system incorporates the K-Nearest Neighbors (KNN) algorithm, Mediapipe library, and a dataset sourced from Kaggle. The KNN algorithm is employed for pose recognition, utilizing the distances between poses to classify and identify the closest match. Mediapipe library is utilized to extract pose landmarks from input video frames, providing valuable information for pose detection. The dataset from Kaggle serves as the training data, enabling the system to learn and recognize various yoga poses accurately. This combination of KNN, Mediapipe, and the Kaggle dataset enhances the system’s ability to perform real-time and accurate yoga pose detection, facilitating effective feedback and guidance for users during their yoga practice. The results obtained from the project demonstrate the effectiveness of the KNN-based system in accurately detecting and recognizing yoga poses in real-time. The accuracy of the system is evaluated using appropriate metrics, providing insights into its performance and ability to assist users in achieving correct poses. The findings of this project contribute to the development of interactive and reliable tools for yoga practitioners, enhancing their practice and improving pose correctness