This demo shows 2 face detectors:

1. OpenCVs Viola Jones implementation (red)
2. YOLOv2 trained on the WIDER FACE dataset (green)

It's using OpenCV's dnn module for YOLO inference. Check out the video below:

You can easily run this demo by cloning this repo and running the run.sh script, or with docker directly:

xhost + && \
docker run --privileged --rm -it \
  -e DISPLAY=unix$DISPLAY \
  -v /tmp/.X11-unix:/tmp/.X11-unix \
   housebw/demo python app.py

The housebw/demo docker image includes all necessary dependencies as well as the trained models.

Check out the 5 pre trained models in the zoo!