| Popis: |
Yoga aims at development of all-round personality, it is synchronization of mind, body and the spirit. Improper postures / in-correct way of doing Yoga can result in serious damage to one’s physique and brain. In ancient time, it used to be performed only under the supervision of an accomplished Guru (teacher). In this busy world with time and place constraints, it is difficult to locate an accomplished Guru. Therefore, it is essential to adopt the right Yoga procedure at the very beginning. Deep Learning models can be trained to detect Yoga postures and be able to provide feedback / corrections if needed. OpenPose was the first real-time multi-person system which brought revolutionary change in the field of pose estimation. In this paper using OpenPose we have created a model which estimates a given human pose. We have compared the results obtained by the model for 2D and 3D points of the image and determined if adding more features to the dataset actually increases the accuracy of the model or not. Furthermore, we have put forward a simple neural network model that efficiently analyses the input image and conveys if the pose performed in the image is correct. |
