Yoga Detection and correction using Mediapipe

"Perfect your pose, Pixel by Pixel"

Comment

Inspiration

With the rise in health-conscious behavior, especially during lockdown, many individuals turned to home workouts and yoga as a way to stay fit often without supervision. Additionally, attending yoga studios or hiring personal instructors can be prohibitively expensive for many middle-class families, making it difficult for them to access proper guidance and posture correction.This reality inspired us to create a Yoga Posture Detection and Correction Website that empowers users to safely practice yoga at home .

What it does

**Yoga Posture Detection and Correction **website that uses real-time pose detection to guide users through proper posture during yoga practice right from their browser in real time.

How we built it

  1. Web Development: Flask (Python)
  2. Pose Detection: Utilized MediaPipe and OpenCV to detect body landmarks via the webcam.
  3. Posture Correction: Compared user poses with ideal reference angles to provide visual feedback ("Correct", "Adjust your back", etc.).

Challenges we ran into

  1. Synchronizing real-time webcam video processing with the Flask server while keeping it performance.
  2. Designing accurate pose-angle-based feedback logic for various yoga postures, which involved trial, error, and a lot of tweaking.

Accomplishments that we're proud of

We’re proud to have developed a real-time yoga posture correction system using just aFlask, MediaPipe, and OpenCV, without relying on complex AI models. The system provides instant feedback based on body joint angles, helping users correct their posture and prevent injuries. It’s simple, privacy-focused, and accessible to anyone with a basic device. Most importantly, it addresses a real-world health issue by making it possible to practice yoga safely at home and at a low cost for individuals who cannot afford personal trainers or high-cost studios.

What we learned

Through this project, we learned how to leverage MediaPipe's 33 body landmarks to accurately calculate joint angles critical to assess yoga poses. We also had hands-on experience with operating MediaPipe and OpenCV together with Flask to enable smooth real-time pose detection and feedback in a web application. We also broadened our understanding of human body kinematics, specifically the movement and alignment involved in yoga, which helped us implement effective posture correction logic.

What's next for

In the future, we plan to add live classes led by yoga experts to provide users with real-time guidance and a more interactive experience. We also aim to offer personalized health tips based on users’ progress and posture data. Additionally, we’re working on integrating audio feedback to make posture corrections clearer and easier to follow, enhancing the overall usability and effectiveness of the app.

Built With

  • mediapipe
  • opencv
  • python(flask)
  • tensorflow
Share this project:

Updates