Yoga Pose Detector

Inspiration

While moving through the planning phase of this year's Knight Hacks, we got on the topic of flexibility and yoga. We also had an interest in challenging ourselves and working with technologies we were not familiar with. At that point, computer vision and motion tracking seemed like an obvious choice.

What it does

The program determines if a person's yoga pose is correct or incorrect based on either the user's webcam feed or a pre-recorded video. The program determines the correctness of a yoga pose by comparing the landmarks of a detected person's body with preset landmark values based on correct yoga poses.

How we built it

Using OpenCV, we were able to pull the web camera's live feed into a computer and offer plenty of basic drawing tools for debugging. Mediapipe was used to then analyze the footage and attach landmarks or points to a body in front of the camera. Python was our programming language of choice

Challenges we ran into

Our team had many ideas they were excited to work on, but with time being a limiting factor and our unfamiliarity with much of what we were working with, we ran into many challenges. Team members Ethan and Devon had very little experience in Python and the whole team had to learn OpenCV and Mediapipe over the course of the hackathon. Also, none of us are particularly good at yoga.

Accomplishments that we're proud of

While many of the ideas we began this hackathon with did not come to fruition, we’re the proud of the time and effort we put into working and learning. We are also proud to have been a part of Knight Hacks and had a working project to submit.

What we learned

Perhaps the most important lesson we learned is that planning is very, very important. Considering the limits of our skillsets and understanding the time it takes to learn new things are all very important aspects to consider when in the planning phase. The whole team also learned a lot more over the course of the hackathon. We all gained experience working with OpenCV and Mediapipe, and Ethan and Devon learned a new programming language.

What's next for Yoga Pose Detector

The main system we were not able to implement in time was a game system where the user was challenged with doing and holding many poses in a row. The would be given a score after several poses and their score would be placed on a leaderboard to be compared with other users.