Yoga Pose Optimization Using Deep Learning Techniques

Yoga Pose Optimization Using Deep Learning Techniques

Jalla Vinay Kumar

Department of Artifical

Intelligence&Machine Learning

MallaReddy University,Hyderabad

Basetti Vinay

Department of Artifical Intelligence&Machine Learning

MallaReddy  University,Hyderabad

Burra Vinayaka Datta

Department of Artifical

Intelligence&Machine Learning

MallaReddyUniversity,Hyderabad

Muthyam Vinesh Goud

Department of Artifical

  Intelligence&Machine Learning

MallaReddy  University,Hyderabad

Abstract—In the digital age, as more yoga practitioners turn to online platforms for instruction, a notable gap has emerged in receiving real-time feedback on yoga poses. Despite the wealth of resources available, such as instructional videos and written guides, learners often struggle to ensure they're executing poses correctly without direct guidance. This lack of immediate feedback can lead to improper posture, potentially causing long- term health issues. Recognizing this challenge, a solution has emerged: automatic assessment of yoga postures using advanced technology.Expanding upon this concept, future iterations could explore additional functionalities, such as personalized feedback based on individual progress, integration with wearable devices for enhanced tracking, and gamification elements to make the learning process more engaging. Furthermore, by leveraging advancements in machine learning and computer vision, this technology could potentially be adapted to other forms of physical exercise, broadening its applications beyond yoga. Overall, the automatic assessment of yoga postures represents a promising advancement in leveraging technology to enhance the practice and accessibility of yoga for enthusiasts worldwide.

However, even after learning or receiving training from the best sources such as videos, blogs, articles or documents, the user does not have time to follow up whether he/she is controlling his/her body properly, and this will be done later. life problems, body posture and health problems. Existing equipment can help in this regard, but yoga practitioners have no way to know whether their body is good or bad without the help of a teacher. Therefore, automatic measurement of yoga poses is aimed at yoga pose knowledge,alerts can be given to practitioners using the Y_PN-MSSD model, in which Pose-Net and Mobile-Net SSD play an important role. This model is divided into three levels. Initially, there is a data collection/preparation phase where yoga poses from four users are recorded and an open file containing seven yoga poses. point for feature extraction training. Finally, yoga poses are recognized and the model helps the user complete yoga poses by tracking them in real time and correcting them instantly with 99.88% accuracy. In comparison, this model outperforms the Pose-Net CNN model. So this model could be the beginning of creating a system that will help people do yoga with the help of a smart, cheap and effective virtual yoga instructor.

Keywords—Yoga, Pose-Net, Mobile-Net, Accuracy,Yoga Pose CNN,Deep Learning