YogaForEveryone

Inspiration

Yoga is known to be an effective way to reduce stress and anxiety. It can help to enhance our mood and the overall sense of well-being. For example, I rely heavily on yoga practice to release my tension after long hours of sitting in front of the computer. Unfortunately, many physical yoga classes have been closed down due to the COVID-19 pandemic. There is no way for people to learn yoga during this period of time. Fortunately, I encountered an apps that can allow me to continue doing yoga from home. I can do yoga using my phone anytime anywhere. This is very good. However, one limitation is that I am not sure whether I am doing the yoga poses correctly because there is no one there to guide me. Therefore, there is a need for a solution that can better improve the yoga practice experience.

What it does

In this project, we propose an automatic yoga pose estimation and analysis system to track and analyse the yoga poses performed by an individual using web camera or camera on smartphone. The system will identify incorrect poses and point out to the user the wrong posture during yoga practice. Then it will highlight the correct way to perform the poses.

How we built it

In this project, we train an AI model using online videos containing the different yoga poses from the Internet. We use altogether 100 videos for training purpose. A pose estimation technique known as AlphaPose was applied to detect the joint position of the human body. After that, we use some pose matching technique to match the pose performed by the user with the standard pose.