About

This project uses Vagrant to spin up a virtual machine running Debian.

It provisions that Debian VM with Docker.

The Debian VM then creates a Debian Docker image which serves a simple website and REST API via the Flask framework for Python.

The website allows users to draw a digit on an HTML canvas.

The website then scales the hand-drawn digit to the size of images from the MNIST dataset of hand-drawn digits.

The scaled image is then base64-encoded and sent to the server to be checked against a neural network model which was previously trained against the MNIST dataset of hand-drawn digits.

The server responds with JSON data containing a prediction and confidence level:

{ 'prediction': '4', 'confidence': '80%' }

The website informs the user of the results.

Setup

Install VirtualBox: https://www.virtualbox.org

Install Vagrant: https://www.vagrantup.com/

Install Vagrant plugins:

vagrant plugin install vagrant-docker-compose
vagrant plugin install vagrant-disksize

Running

vagrant up

^ this will take awhile the first time

Verify things are working

Test POSTs:

./scripts/test_post_zero.sh

./scripts/test_post_three.sh

Test GET index in browser:

http://192.168.188.110:8081

List running Docker containers from the VM:

vagrant ssh
docker ps

Maybe get shell on a container through the VM:

vagrant ssh
scripts/bash_docker.sh $CONTAINER_ID

Automatically sync files between host and VM

vagrant rsync-auto

^ this command watches all files in the host machine's top-level project directory (as specified by the config.vm.synced_folder line in the Vagrantfile)

if a change is made, an automatic rsync will update files in the VM's /vagrant/ directory, thus ensuring that both directories (the host machine's and the VM's) contain the same content.

vagrant rsync can be used for a one-off file sync

Resources

Spinning up a VM with Ubuntu and Docker, via Vagrant

https://technology.amis.nl/2018/05/21/rapidly-spinning-up-a-vm-with-ubuntu-and-docker-on-my-windows-machine-using-vagrant-and-virtualbox/

Simple Docker container + Flask server

https://blog.ouseful.info/2018/08/06/docker-container-and-vagrant-virtualbox-101s/

Flask server with simple Machine Learning REST API

https://medium.com/@dvelsner/deploying-a-simple-machine-learning-model-in-a-modern-web-application-flask-angular-docker-a657db075280

React + Redux with Flask backend

https://github.com/YaleDHLab/flask-react-boilerplate

Docker Compose

https://docs.docker.com/compose/gettingstarted/

Faster builds

https://medium.com/google-cloud/deploying-scipy-numpy-with-docker-205e9afac3b0

Basic classification NN with Tensorflow

https://github.com/tensorflow/docs/blob/master/site/en/tutorials/keras/basic_classification.ipynb

Saving, loading, and serving Tensorflow models

https://colab.research.google.com/github/tensorflow/docs/blob/master/site/en/tutorials/keras/save_and_restore_models.ipynb?hl=es-MX

User-drawn MNIST images

https://codepen.io/MaciejCaputa/pen/KXbJWR

Deep Convolutional Generative Adversarial Networks with TensorFlow

https://github.com/dmonn/dcgan-oreilly/blob/master/DCGANs%20with%20Tensorflow.ipynb