pull-up tracker

• a custom trained object detection YOLO (v. 5) model that can recognize people doing pull-ups with excellent results:
• a script (counting_single.py) that, given a video of one person training, can live count how many repetitions are being done
• a script (challenge.py) that tracks (with custom developed class Tracker()) how many pull-ups are executed by each person in the video, as in a competition

Usage

git clone https://github.com/davide710/pull-up_tracker.git
cd pull-up_tracker
pip install numpy opencv-contrib-python pyyaml

# To execute a script with a video
python3 [script.py] /path/to/video
# To execute the script challenge.py with the example
python3 challenge.py videos/pullups_challenge.mp4
# To execute it with live webcam (you may need to adjust webcam configurations in the script)
python3 [script.py] -live

This is the result obtained with videos/pullups_challenge.mp4:

The original video was taken from https://www.youtube.com/shorts/ft7VmEyvcuc?feature=share

"Manually counting repetitions? Yeah, I think the ancient Greeks used to do it too"