- models.py, utils.py and test.py are taken (and adjusted) from oarriaga
- I am using the pre-trained models from oarriage, which he trained on the
fer2013.bibdataset - Image face-classification is done by opencv's haarcascade
- oarriaga's project is under MIT License
NOTICE: call these from the project root, not from inside the ./face-recog directory
./face-recog/prepare.sh
# installs python dependencies via pip
# also make sure you already have built the docker-image for the XOR project /README.md#1
python face-recog/export.py # make sure to run from project root folder
# results will be in (/export)
./start-dual-servers.sh
# ./stop-dual-servers.sh
- the client loads an image from whatever path (relative/absolute) you pass as first argument
cd ./face-recog/client
yarn (npm install -g yarn, in case you do not have it installed already)
yarn start ./../images/cutouts/4.png
# make sure to start the servers first
cd ./face-recog/
python face-exporter.py ./images/chris_1.jpg
# makes a gray cutout in 48x48 pixel of a face in the image
cd client
yarn start ./../images/face-exports/gray_0.png