Yoga Pose Detection using Machine Learning

Abstract

Yoga is a healthy way of life, originated in ancient India it has its followers growing all around the globe. With India as its epicenter gurus are the guiding light in practicing yoga efficiently and most importantly in a traditional way in which it was meant to be. As time went by, we faced a global pandemic which impacted millions of lives and instilled fear in the hearts of the general public limiting them to their homes with their loved ones. This affected the public gatherings for yoga, yoga classes which forced people to practice it at their respective homes, which led to a lack of mentorship and posture emphasis among the practitioners. To enable them and also help them in practice this model was prepared so that practitioners can see themselves in real-time performing different postures and also test themselves that if they are doing it rightly or not. With the help of mediapipe, a deep learning architecture was prepared so that landmarks of the human body can be processed correctly for recognition and testing. The model is trained with various yoga - postures increasing its accuracy and faithfulness among users.