To run you need to have a few dependencies installed (I recommend installing everything in a virtualenv):
- dlib python (available in pip)
- opencv 3.2 (compile from source with python support recommended)
- numpy
- sklearn
- matplotlib
- IPython (for data collection)
- pandas
- pickle
- Python >= 2.7.10
Start the server (while in your virtualenv) using:
python src/smile_server.py
If matplotlib complains a bunch on OSX, you should read http://matplotlib.org/faq/osx_framework.html
You should also probably create the directory data/ as soon as you pull the repo.
data/smile_home is where faces and identities received through the api point /train will get saved. So take good care of it.
The first time you run the program, the LFW (face database) will be downloaded and it could take a long time.
Preprocessing images are cached so future runs of the app should be quicker than your first time.