Implemented and optimized Blaze Pose, a lightweight convolutional neural network architecture, for human pose estimation on mobile devices. Investigated the application of L2 Normalization to rendered coordinates in pose estimation to ensure consistency and accuracy in vector representation. Developed a method for Yoga Pose Assessment using Blaze Pose and DTW to provide real-time feedback on pose correctness. Achieved a frame rate of over 30 frames per second for pose estimation, making it suitable for real-time applications such as fitness tracking. Proposed a novel approach for Yoga Pose Recognition targeting both beginners and experienced specialists. During this research gained strong understanding convolutional neural networks for pose estimation proficiency in TensorFlow, and OpenCV. Above work was presented to the examiners during the course of Bachelor’s of Engineering.