Real-time CORRECT Form Push-Up Counter

This project helps track and improve push-up form by detecting repetitions and correctness in real-time.
I built it to learn computer vision - and also to improve my own push-ups (I'm using this program daily)

Notes

• The program counts how many correct push-ups you do in real time
• A push-up is considered correct if:
  • the elbow angle goes below 90° when going down
  • and it exceeds 145/155° when going up
• At each step the program tells you when to go down / up, so that you are on the correct pace.
• You could also make the push-ups harder by adjusting 'valid_frames' value (from 1 to 5 / 6 or 10).
  • The higher the value, the longer you must hold the “up” or “down” position.
• Also if you want to reset the counter you can just clap once and it resets to 0.

Tools Used

• Python
• MediaPipe - i'm using only 3 * 2 keypoints : left/right shoulder, elbow and wrist
• OpenCV - camera input
  • note that my camera is only 1280x720, if yours is 1920x1080 adjust the resolution : CAMERA_WIDTH / CAMERA_HEIGHT

How to Run

1. Clone the repository:

   git clone https://github.com/RadwB2112/RealTimeCorrectPushUpCounter.git
   cd RealTimeCorrectPushUpCounter  

2. Install the requirements

   pip install -r requirements.txt

3. Run the program and start doing push-ups )))

   python3 main.py