Inspect the Dockerfile. It puts together an Openface installation, with all dependencies, on an Ubuntu system, along with a REST app.
Change current folder to spektor-brain
Issue:
make docker-build
This will build a docker image spektor/brain. It can take significant time because some dependencies are built from source.
In the meantime, download models listed in file urls.txt inside spektor-brain/src/service/data folder. These are pretrained neural networks used by Openface provided by authors and maintainers of Openface and DLib.
I encourage to use docker installation instructions from Digital Ocean https://www.digitalocean.com/community/tutorials/how-to-install-and-use-docker-on-ubuntu-16-04
After the image is built you can run with with
make docker-run
Issue docker logs spektor to see if all went well (loading models upon image start takes few seconds).
In the logs you should see something like this:
INFO:spektor:Starting app INFO:spektor:Models loading.. INFO:spektor:Done. INFO:spektor:App ready! INFO:werkzeug: * Running on http://0.0.0.0:5000/ (Press CTRL+C to quit)
The app works on port 5000 and exposes two endpoints. One is simple hello, the other is identify and is used for finding faces on a JPEG photograph.
With spektor/brain running you can start the backend REST app. Change current folder to spektor/backend.
First you need to install required dependencies (see requirements.txt).
To install them issue:
pip3 install -r requirements.txt
Once this this is completed issue:
./spektor-cli.py
This will launch the second REST app. It will work on port 6900, communicating with the dockerized app and the database (sqlite 3 based).
There are two kinds of enpoints this app exposes. The first group enables interaction with the database. These are setup with flask-restless and allow creation, modification and deletion of entries in the database. On top of that there is flask-admin setup, which enables database management through a browser. To see it in action open
http://localhost:6900/admin/
The second group of endpoints is responsible for handling requests to analyze the submitted image and execute matching algorithm.
There are two demonstration command line scripts written in python which allow to track what kind of interactions can be made with the REST endpoints and what the workflow could look like.
First of them, add.py, could be used to populate the database with faces found on new photographs.
Run ./add.py --help to see available help.
What it does is it finds faces in specified photographsm picks the most probable face, analyzes it and adds coresponding data to the database (new persona, with a thumbnail and associated image).
You need to either specify personal data from the command line or to enable interactive mode to enter them using keyboard.
The second script, analyze.py, takes given photographs, finds faces, analyzes them, executes match against profiles in the database and prompts for action.
Face can either be (1) skipped (e.g. falsely detected face), (2) used to create new persona (Joe Doe), or (3) added as another representation of a person's profile in the database (one person can be assigned more than one face to be matched against).