A COMPARATIVE STUDY OF DEEP LEARNING AND ENSEMBLE METHODOLOGIES WITH KEYPOINT DETECTION FOR 2D IMAGE-BASED YOGA POSE RECOGNITION

Abstract

Yoga is a 5000 year old Indian spiritual practise that seeks to achieve balance between the body and mind, with the use of asanas, meditation, plus a variety of breathing exer cises/techniques. Yoga is now widely used as a remedy for dealing with the rising levels of tension and anxiety while adjusting to modern lifestyles. We can reap the greatest health benefits from yoga by adopting the ideal postures and adhering to the recommended techniques and sequencing. A multitude of techniques can be used to learn yoga, such as through attending courses at yoga studios, viewing films, perusing photographs, or reading books. The majority of people prefer self-learning at home due to their fast-paced existence. However, many find it challenging to see defects in their own yoga positions. However, adopting improper postures while practising yoga can result in a number of health issues, including short-term chronic issues and acute muscle discomfort. Therefore, there is a need for scientific evaluation of incorrect yoga posture detection and correction through providing feedback in order to help people practise yoga effectively. To support self-learning, we’ll provide a model for classifying poses using posture detection in this research. Users will pick a yoga stance to practise here and upload a picture or a video of it. The user pose is supplied to training models, which figure out and output the differences between the user pose’s body angles and the real pose that was detected. With this output, the model instructs the user on how to enhance the position by pointing out the areas of their yoga posture that need work.