Yoga Pose Correction App

Aligned is a yoga pose correction web application that uses computer vision and deep learning to process videos of a user’s yoga poses and provides feedback on how to improve the pose. The app was built using Flask and runs on an Amazon EC2 instance with GPU.

How does it work?

• User logs in and selects which pose to evaluate
• App provides instructions on how to execute the pose
• Once ready, the user records a video of themselves doing the pose within the app (using webcam)
• Video is automatically processed using OpenPose, an open-source, deep learning library for keypoint detection developed at Carnegie Mellon University.
• The app uses these keypoints from OpenPose as features to classify different pose faults using a rule-based algorithm.
• Feedback is provided to the user on how they can safely improve their pose.

Note that the app's website is not currently available due to the high cost of running AWS EC2 instance.

🎥 👉 Watch Demo Video

As a certified yoga teacher, I was thrilled to combine my two passions in this project: data science and yoga.

Main Scripts:

• Slide Deck for final presentation to VCs
  
• Flask routes
• Video processing script
  
• Modeling