Hey, in this post we’ll build a web application that can handle images taken using our mobile app that we’ve previously built and store them in the database. For sake of simplicity I decided to do this using Django & Django-REST-framework.

After we’re done with this, we’ll perform image processing (in the next post) and build a predicive model.

By now, we only have base64 images that cannot be saved in the database yet. But, what is base64 and why we cannot save images represented in this format in the database?

What is base64 format?

Base64 is a way to represent binary data (stream of 1s and 0s) in an ASCII string format.
It was introduced in order to avoid data corruption when transmitted in binary format across a network.

Examples of binary data corruption:

  • Some protocols may interpret your binary data as control characters (like a modem).
  • The underlying protocol might think that you’ve entered a special character combination (like how FTP translates line endings).

So, after base64 image was successfully transmitted we can decode it back to binary data.
There are two reasons why we are doing that:

  • Later on, we want to perform image processing using scikit-image and opencv both of which naturally only support binary input.
  • Django ImageField requires image in binary format. ImageField is an object that lets you save image files in the database.

Now, let’s talk a little bit about the main topic of this post.

What is Django?

Django is a free and open source web application framework, written in Python. The role of Django boils down to communicating with your app’s database and serving particular data to client along with static files. Also, it has a set of handy components that let you for example handle media uploads, form validation or user authentication.

What is Django REST Framework?

Django REST Framework(DRF) is a third-party app for Django. DRF gives you a lot of convenience in inspecting and managing you database(via “Browsable” API).


Just to recall, we are sending cropped base64 image over the http protocol just like this:

// Specify appropriate server endpoint address
// Ending slash in the address is necessary
        // Data passed along a request
        image: cropped_img_base64
    function(data, status, xhr) {
        // Success callback
        alert('Status: ' + status + '\Data: ' + data);
.fail(function(error, status, xhr) {
    // Failure callback
    alert('Status: ' + status + '\nReason: ' + xhr);

We use $.post method which performs asynchronous HTTP requests to our server.


Before we dive into building a web application, let’s go through project setup first.

DRF Setup

  • Install python 3.5:
    Windows x86-64 web-based installer or Select & download other installer

    While installing Python make sure that option ‘Add Python 3.5 to PATH’ is checked.

  • Create a new Django project named digits_operators_recognizer, then start a new app called resolver:

    # Install virtualenv package
    # pip is a package management system used to install and manage software packages written in Python
    pip install virtualenv
    # Create the project directory
    mkdir digits_operators_recognizer
    cd digits_operators_recognizer
    # Create a virtualenv to isolate our package dependencies locally
    virtualenv env
    source env/bin/activate     # On Windows use `env\Scripts\activate`
    # Install Django and Django REST framework into the virtualenv
    pip install django djangorestframework
    # Set up a new project with a single application
    django-admin.py startproject digits_operators_recognizer .    # Note the trailing '.' character
    cd digits_operators_recognizer
    django-admin.py startapp resolver
    cd ..
  • Add ‘rest_framework’ to your INSTALLED_APPS setting:



Basically, building any Django project requires the same set of steps you need to take. This set of steps includes:

  • Define Django models and fields.

    Generally, each model represents a single database table, e.g. users table.
    Fields are table columns, e.g. username, password.

    # digits_operators_recognizer/resolver/models.py
    # ...
    class Image(models.Model):
        # Define fields like below
        image = models.ImageField(upload_to='images/%Y/%m/%d/')
        # auto_now_add set to true makes this field filled with the date of entry creation
        timestamp = models.DateTimeField(auto_now_add=True)
        # we will add more fields here later
        # Django is using this method to display an object in the Django admin site
        def __str__(self):
        	 return self.image
  • Specify URL address and point it to a particular view.

    Before users can visit a website, you must specify available URL addresses.

    # digits_operators_recognizer/urls.py
    # ...
    from django.conf.urls import include
    from django.conf import settings
    from django.conf.urls.static import static
    urlpatterns = [
        # Here we include resolver app urls
        url(r'^api/', include('digits_operators_recognizer.resolver.urls')),
        # Address to admin site
        url(r'^admin/', admin.site.urls),
    # It lets us preview uploaded images through Browsable API
    if settings.DEBUG:
        urlpatterns += static(settings.MEDIA_URL, document_root=settings.MEDIA_ROOT)

    Django project consists of apps. Each of these apps are supposed to be kept completely separate from each other. That’s why they have their own set of models, urls and views.

    # digits_operators_recognizer/resolver/urls.py
    # ...
    from digits_operators_recognizer.resolver import views
    from rest_framework.urlpatterns import format_suffix_patterns
    # Here we specify available URL addresses
    urlpatterns = [
        url(r'^images/$', views.ImageList.as_view()),
        url(r'^images/(?P<pk>[0-9]+)/$', views.ImageDetail.as_view()),
        url(r'^images/create/$', views.ImageCreate.as_view()),
    # Adding this lets you use filename extensions on URLs to provide an endpoint for a given media type.
    # For example you can get endpoint data in json representation or html static file
    urlpatterns = format_suffix_patterns(urlpatterns, allowed=['json', 'html'])
  • build views

    View defines what will happen after someone enters corresponding URL address.

    # digits_operators_recognizer/resolver/views.py
    import os
    from django.conf import settings
    # Our model
    from digits_operators_recognizer.resolver.models import Image
    # Our serializer
    from digits_operators_recognizer.resolver.serializers import ImageSerializer
    # DRF modules
    from rest_framework import status
    from rest_framework.request import Request
    from rest_framework.response import Response
    from rest_framework.generics import ListAPIView, RetrieveAPIView, CreateAPIView
    from rest_framework.permissions import IsAdminUser
    # OCR scripts
    from ocr import image, recognizer
    class ImageList(ListAPIView):
        'List all images'
        serializer_class = ImageSerializer
        permission_classes = (IsAdminUser,)
        queryset = Image.objects.all()
    class ImageDetail(RetrieveAPIView):
        'Retrieve an image instance'
        serializer_class = ImageSerializer
        permission_classes = (IsAdminUser,)
        queryset =  Image.objects.all()
    class ImageCreate(CreateAPIView):
        'Create a new image instance'
        serializer_class = ImageSerializer
        def post(self, request):
            serializer = ImageSerializer(data=request.data)
            if serializer.is_valid():
                # Save request image in the database
                return Response(serializer.data, status=status.HTTP_201_CREATED)
            return Response(serializer.errors, status=status.HTTP_400_BAD_REQUEST)

Last but not least, I would like to introduce you to DRF serializers.

Serializers allow complex data such as querysets and model instances to be converted to native Python datatypes that can then be easily rendered into JSON, XML or other content types.
Serializers also provide deserialization, allowing parsed data to be converted back into complex types, after first validating the incoming data.

In other words, a serializer translates given queryset to human readable format and vice versa.

For sake of clarity, I decided to create a separate file named serializers.py inside resolver directory where we will declare our serializer.

# digits_operators_recognizer/resolver/serializers.py

from digits_operators_recognizer.resolver.models import Image
from rest_framework import serializers

from django.core.files.base import ContentFile
import base64
import six
import uuid
import imghdr

class Base64ImageField(serializers.ImageField):

    def to_internal_value(self, data):

        # Check if this is a base64 string
        if isinstance(data, six.string_types):
            # Check if the base64 string is in the "data:" format
            if 'data:' in data and ';base64,' in data:
            	# Break out the header from the base64 content
            	header, data = data.split(';base64,')

            # Try to decode the file. Return validation error if it fails.
            	decoded_file = base64.b64decode(data)
            except TypeError:

            # Generate file name:
            file_name = str(uuid.uuid4())[:12] # 12 characters are more than enough.
            # Get the file name extension:
            file_extension = self.get_file_extension(file_name, decoded_file)

            complete_file_name = "%s.%s" % (file_name, file_extension, )

            data = ContentFile(decoded_file, name=complete_file_name)

        return super(Base64ImageField, self).to_internal_value(data)

    def get_file_extension(self, file_name, decoded_file):

    	extension = imghdr.what(file_name, decoded_file)
    	extension = "jpg" if extension == "jpeg" else extension

    	return extension

class ImageSerializer(serializers.HyperlinkedModelSerializer):
    image = Base64ImageField(
        max_length=None, use_url=True,

    class Meta:
        model = Image
        fields = ('pk', 'image', 'timestamp')

Running server

# Sync the database - propagate changes you made to your models into your database schema
python manage.py migrate

python manage.py runserver

Full code

Visit my repo: digits_operators_recognizer


In the next post we will perform image processing in order to extract relevant features that can be fed into predicive model.