git clone [email protected]:husnoo/skynet.git

https://skynet999.slack.com/messages/@slackbot/

https://docs.google.com/presentation/d/1aWrTjkXcW9VfOehtvR04VAuFhb6OkSsHRSy3sFIyxOY/edit#slide=id.p

Global Hackathon 2017 London

Top level Documentation

The working code is under app/.

There are a few requirements for development and deployment:

1. A valid SSL certificate - we use a self-signed one for development, but this will give a warning/error on most browsers.
2. A valid API Key from Microsoft for the Face API - note we give this away to the user. For development, either get a free one with their dev trial, or hackathon giveaway pass, or just don't bother using one, in which case the student code will return random numbers.
3. A machine running docker

The backend code lives in app/skynet.py, and app/get_score.py. The latter is the neural network that characterises the Face API output for our purpose, and the training data lives in app/theta1.txt and app/theta2.txt.

app/skynet.py is the main server code. It exposed the endpoints:

@app.route('/static/js/student.js')
@app.route('/static/<path:path>')
@app.route('/teacher')
@app.route('/send_emotions', methods=['POST'])
@app.route('/teacher-emotions')
@app.route('/')

1. The first one is for student.js to have the right API key. Terrible design of course, giving away your key.
2. The second is for any js/css/etc files.
3. /teacher shows the teacher web page, by reading 'app/html/teacher.html'.
4. /teacher-emotions returns a dictionary of values for each connected user.
5. / is the main page, which is aimed at students, so it reads in 'app/html/teacher.html' and serves that.
6. /send_emotions is a POST endpoint, where students send their emotions as returned by the MS Face API.

Other trivia

1. We're using the Flask web library, which handles the boring bits of HTTP back and forth, as well as wrapping it all in a HTTPS layer. This is needed for the client camera to be available to us.
2. We give people a cookie, not because we're nice, but because we want to know who's paying attention.
3. get_data and set_data store the application state, i.e. who's listening, and how much attention they're paying. In an operational setting we'd use redis.

Building the docker image

To use the self-signed certificate and invalid MS Face API Key:

git clone [email protected]:husnoo/skynet.git
cd skynet/app
cp ./keys/demo.keys/host.crt ./keys/host.crt
cp ./keys/demo.keys/host.key ./keys/host.key
cp ./keys/demo.keys/azure_face_api_key.txt ./keys/azure_face_api_key.txt
docker build -t skynet .

To use proper credentials for live deployment:

1. Go to https://portal.azure.com/, create a "Cognitive Services" service.
2. Under "Resource Management", look for Keys, and copy out "Key 1" and store it in a file called azure_face_api_key.txt.

git clone [email protected]:husnoo/skynet.git
cd skynet/app
cp path/to/your.own.keys/somename.crt ./keys/host.crt
cp path/to/your.own.keys/somename.key ./keys/host.key
cp path/to/your.own/azure_face_api_key.txt ./keys/azure_face_api_key.txt
docker build -t skynet .

Running locally

docker run --rm -ti -p 0.0.0.0:443:443 --name skynet skynet

Go to:

https://127.0.0.1/

To stop it:

CTRL+C

Deployment

docker save skynet --output skynet.docker
tar -cvzf skynet.docker.tar.gz skynet.docker
scp skynet.docker.tar.gz [email protected]:/tmp/skynet.docker.tar.gz

On ownserver.com:

tar -xvzf /tmp/skynet.docker.tar.gz -C /tmp/
docker load --input /tmp/skynet.docker
docker run -d -p 0.0.0.0:443:443 --name skynet skynet &

To stop it:

docker rm -f skynet

To regenerate a self-signed certificate

cd skynet/app/keys/demo.keys/
openssl req -x509 -sha256 -nodes -days 365 -newkey rsa:2048 -keyout host.key -out host.crt

Clean docker image cache

docker images -q |xargs docker rmi

Note that sometimes the browser can also cache .js files, so your changes won't show. Use Incognito Browser for better test results.