Yoga Posture Analysis using Deep Learning
Sagar Wadhwa, Anshumaan Garg, Geetika Munjal · 2024
We live in a society where happiness is crucial for living and this happiness can be attained by having a fit body. For this fit body, yoga is a key element. Yoga posture analysis using deep learning is a significant step for fusing ancient wisdom with the newfound cutting edge and innovative technology. Specialized deep learning methods have been used such as CNN (Convolutional neural network). It is the main neural network for analyzing the yoga poses. This ability is due to the structure of CNN, the intricate patterns it can recognize in the data. These models are also a way forward in the case of image recognition. Key alignments are carefully identified and analyzed with respect to the posture. Initial results have shown YOLO to have an average precision of 92.5% on the kaggle dataset, as compared to 91.80% precision for MediaPipe and 85.94% for OpenPose. All the models discussed in this paper are well capable of recognizing poses and layers of network are working here for determining the posture.