Deep learning model deploy with Django


Executive Summary

This project is a Django-REST API that offers the consumption of a deep learning model using a simple front end. The model adopted in this work is the previous version of an Emotion Classifier trained with audio files of the RAVDESS dataset. To know more, see this repository.

How does this work?

User Journey

The user journey starts on the index page at /index/ where it is possible to choose if

1) Upload a new file on the server; 2) Delete a file from the server (WIP); 3) Make a prediction on a file already on the server;


Choosing Upload your audio file the user will be redirected to a modified home page. The user will be asked to pick a file from his computer. The UI will confirm if the operation has been successful.


Choosing Make your prediction the user will be redirected to a modified home page. In this page, it will be possible to see a list of the files already on the server. Following the path media/{filename} it will be also possible to listen to the audio file.


After clicking on Submit, the user will be redirected to a modified home page that will include the prediction made by the Keras model for the file selected.


See the App in action!

There is a short demo of the first version on YouTube: . The UI has been updated since then, as now manages all the actions extending the index templates with the action templates. The above pictures are updated with the new workflow.

The current version of the application is deployed on Heroku: . Please note the performances are not good as the machine used is a free tier.

Developers stuff

DB creation

PostgreSQL needs to be installed. To facilitate the configuration, I suggest to use a Db manager with UI, like pgAdmin.

After the installation of PostgreSQL, it is possible to use pgAdmin to create a django-emotion-classification database and a App_filemodel table.

The App_filemodel table can be created with the following script:

CREATE DATABASE django-emotion-classification;


CREATE TABLE App_filemodel (
   file TEXT NOT NULL,
   timestamp DATE NOT NULL,

GRANT SELECT, INSERT, UPDATE, DELETE ON ALL TABLES IN SCHEMA django-emotion-classification TO marco;

ALTER USER marco CREATEDB; -- This is to run the automatic tests, otherwise you will get an "unable to create database" error when running python test

Please note the above script is made with the data available in the, but it is possible to change it if needed.

    'default': {
        'ENGINE': 'django.db.backends.postgresql_psycopg2',
        'NAME': 'django-emotion-classification',
        'USER': 'marco',
        'PASSWORD': 'test',
        'HOST': 'localhost',
        'PORT': '',
        'OPTIONS': {'sslmode': 'disable'},

How to start the server and try it

1) git clone 2) $ pip install -r requirements.txt 3) Open a terminal window, cd into the project folder and run python runserver.

How to run the tests

python test

Other important topics

The Keras model is stored in the models folder.

gitmedia folder includes the pictures used for this README.

media folder includes the audio files loaded using the server.

It is possible to have an overview of the application even without knowing how Django works by looking at the App/ file.

User Stories