AI Yoga Gesture Detection

Abstract

The purpose of this project is to develop a joint point analysis-based AI-powered yoga posture detection system. The main goal is to create a virtual trainer that can accurately recognize different yoga poses and provide users with immediate feedback. The technology employs sophisticated computer vision algorithms to detect the user's stance by analyzing key joint locations in their body and then advising them on how to correct their posture. The AI yoga gesture detection model achieved an impressive overall accuracy of 95% during training, demonstrating its ability to learn from the dataset and make accurate predictions. When tested on the testing dataset, the model maintained a high accuracy rate of 90%, indicating strong performance in classifying yoga poses on previously unseen data. However, a validation accuracy of 60% indicates a discrepancy between the model's performance on the testing set and its generalization ability. Despite this, the model's high overall accuracy during the training and testing stages demonstrates its ability to accurately identify yoga poses, assisting users in achieving proper alignment and form during yoga practice.