Inspiration

As passionate gym enthusiasts, we often face several issues during our lifts in the gym. One such lift that needs form correction is squats. This project is inspired from that idea where we can leverage computer vision to correct our squat's form.

What it does

Takes in a video input or live video and gives feedback as an overlay on top of video. The feedback provided on top of the video can give recommendations such as:

  • Knees falling over Toes
  • Bend Forward
  • Bend Backward
  • Squat too deep

Number of correct reps and incorrect reps is also calculated. There is beginner and pro modes for which the analysis can vary.

How we built it

Languages used: Python

Libraries used: NumPy, OpenCV, MediaPipie, Streamlit

Challenges we ran into

We faced into multiple challenges throughout the course of development. Some of the notable challenges that we faced are:

  • Frontend Streamlit errors: Uploading video, Loading livestream
  • Get landmark features function: Implementing methods to receive coordinates of shoulders, nose, hips, knees, ankles and knees coordinates
  • Finding angle between knees and feet
  • Libraries and virtual environment configuration issues as well

Accomplishments that we're proud of

We are proud of building a functional app that solves the need we intended to solve. Although the app is only built using Python, we used a wide range of libraries to implement the features in the app.

What we learned

We learnt many new libraries in Python such as MediaPipe, OpenCV, Streamlit and some amount of math using NumPy throughout the development process. The most important thing that learnt is to persist through errors and take things one at a time to solve problems.

What's next for Squat form Analysis

Some future ideas for Squat form Analysis is to implement a Deadlift form analysis program and further extend it to all possible exercises in the gym. Voice features can also be implemented to provide feedback during the reps.

Built With

Share this project:

Updates