Abstract- This paper presents a method for reliably recognising various Yoga asanas using deep learning techniques. A dataset of six Yoga asanas (Bhujangasana, Padmasana, Shavasana, Tadasana, Trikonasana, and Vrikshasana) was constructed using 15 people (ten men and five women) with a standard RGB webcam. A Logistic Regression and long short-term memory (LSTM)-based hybrid deep learning model is suggested. (LSTM) for real-time video yoga detection, using a LR layer extracting features from keypoints of each pose. OpenPose frames are used to generate temporal predictions, which are then followed by LSTM.