Yogasana classification using Deep Neural Network: A Unique Approach

Abstract

Yogasana, often simply referred to as “asana”, is a Sanskrit term that encompasses the physical postures or poses practiced in yoga. Traditionally, there are almost 90 classical Yogasanas, each with its unique benefits and variations. It's very important to identify the postures and benefits of each asana. Classifications of Yogasana can have a significant impact on the practice and promotion of yoga, making it more accessible, educational, and safe for people of all skill levels. In this research work, we have automatically identified different Yogasanas from images using a deep neural network. For our work we have considered 10 very popular categories of Yogasana and built a dataset of volume 800 images. Furthermore, the data augmentation (DA) technique is utilized to increase the volume of data by including slightly modified replicas of existing data. Xception, combined with a self-attention network, has been employed to classify various Yogasanas based on diverse images. The model attains an impressive validation accuracy of 99%, surpassing the performance of all other contemporary state-of-the-art models.