facemaskdetect

Face masks detector with Tensorflow Keras and MobileNet V2. It supports still image or camera detection. The output is to image with bounding box and label (Mask, NoMask), image viewer or to JSON.

Disclaimer

Source code adpted from Adrian Rosebrock article available here: https://www.pyimagesearch.com/2020/05/04/covid-19-face-mask-detector-with-opencv-keras-tensorflow-and-deep-learning/

How to build

To build the docker image please do:

docker build . -t facemaskdetect

How to install

We provide a python setup to install a command line script facemaskdetect:

python setup.py install

How to run

To run the installed script to get JSON output:

facemaskdetect -i examples/example_01.png -o json

for image output with bounding boxes:

facemaskdetect -i examples/test.jpeg -o out.png

How to run the container on linux

To run on a linux host without X server support:

docker run --rm -it -v $(pwd):/app facemaskdetect bash

to enable X server support:

docker run --rm -it --net=host --ipc=host -e DISPLAY=$DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix --env="QT_X11_NO_MITSHM=1" -v $(pwd):/app facemaskdetect bash

How to run the container on macOS

To run on macOS with X server support enabled, first be sure to have xquartz and socat installed:

brew install socat
brew cask install xquartz

NOTE. In case of errors, please do brew reinstall.

then open XQuartz server and bind to the port:

open -a XQuartz
socat TCP-LISTEN:6000,reuseaddr,fork UNIX-CLIENT:\"$DISPLAY\"

then in another window (important!) run docker with display forwarding:

export IP=$(ifconfig en0 | grep inet | awk '$1=="inet" {print $2}')
xhost +$IP
xhost +local:docker
docker run --rm -it -e DISPLAY=$IP:0 -e XAUTHORITY=/.Xauthority --net host -v /tmp/.X11-unix:/tmp/.X11-unix -v ~/.Xauthority:/.Xauthority  -v $(pwd):/app facemaskdetect bash

Mask detection from still image

To detect from a still image with camera output

python facemaskdetect/detector.py -i examples/example_01.png -o cam

To detect from a still image with file output python facemaskdetect/detector.py -i examples/example_02.png -o output.png


To detect from a still image with `JSON` output
```bash
python facemaskdetect/detector.py -i examples/example_01.png -o json

To JSON output format looks like:

[
    {
        "label": "Mask",
        "accuracy": "0.93",
        "box": {
            "start_x": 173,
            "start_y": 161,
            "end_x": 565,
            "end_y": 750
        }
    },
    {
        "label": "Mask",
        "accuracy": "0.99",
        "box": {
            "start_x": 394,
            "start_y": 411,
            "end_x": 685,
            "end_y": 775
        }
    }
]

Mask detection from video

To continuosuly detect from a video stream of a camera (like web cam):

python facemaskdetect/camera.py

Sample application

A example application is available in app.py. To run it

python app.py

loretoparisi / facemaskdetect