Sayuri evaluates good or bad of your conference by machine learning.

• Use web camera to take pictures of conference scene.
• Detect faces in the picture by Rekognition API.
• Evaluate good or bad of conference by machine learning (SVM).

more detail is described below.

machine learning for conference consulting system

• Python 3.4.3
• Tornado
• Redis
• scikit-learn

You can read how to create the model by ipython notebook (at sayuri/machine/evaluator/create_evaluator). To read this document, you have to install below.

• ipython notebook
• matplotlib

You can use Heroku Button.

Sayuri uses Rekognition API, so you have to create account and set configuration variables.

You have to create Rekognition account too. And then create envs.json at the root of the project like below.

{
  "SECRET_KEY": "__YOUR_SECRET_KEY__",
  "REDIS_URL": "redis://localhost:6379",
  "FACE_API_KEY": "your_key",
  "FACE_API_SECRET": "your_secret_key",
  "FACE_API_NAMESPACE": "namespace",
  "FACE_API_USER_ID": "user_id"
}

And You have to create a certificate to use SSL. Because Sayuri use wss protocol for security issue.

Create ssl folder at the root of the project and make serverkey.pem and servercrt.pem.

For example.

openssl genrsa -out serverkey.pem 2048
openssl req -new -key serverkey.pem -out server.csr
openssl x509 -req -in server.csr -signkey serverkey.pem -out servercrt.pem