Project to demonstrate machine learning using Tensorflow python library.
For this project I've decided to build image uploading service using Flask-RESTful API written in Python. Any client application that can make AJAX/REST requests may upload an image to a service using HTTP protocol and JSON DTO (Data Transfer Object). In response, service promises to "study" images and fulfilling it by returning results of it's learning outcome.
What service learn from images we send to it?
We will send MNIST handwritten digit images to it and see if it can return numeric representation of it back to the client application. I've converted binary data sets into png format images in this subproject. It is a good idea to add canvas on the client side to allow user to draw and send image snippets to the service for digits recognition.
How service learn from images?
Tensorflow library is a tool to drive service learning components. Tensorflow developers implemented for us complex mathematical functions required for programs to be able to learn from data we feeding it with. To build app faster, I will use Keras library built on top of Tensorflow. It provides convenient way to create neural network model for training and testing image sets.
What is Flask used for in this project?
Flask provides infrastructure for REST Aware services. One of the key components of Flask-RESTful API are the Resources defined as a classes that contain HTTP methods get, post, put, delete, etc.
https://flask-restful.readthedocs.io/en/latest/quickstart.html
Extensions:
- flask cors -
pip3 install flask-cors- to support cross origin requests
set FLASK_APP=main.py
flask run
What is Tensorflow used for in this project?
Tensorflow is a scientific library for studying data and teaching programs to learn data patterns. When program is trained well enough, it can be used to predict data patterns that are not yet part of the training set.
It can be used to build neural network, simillar to our brain neurons wired together.
In my project, I am using Keras library that is built to use Tensorflow in more convenient way. It is still requires Tensorflow as a backend to run.
- Installing Tensorflow -
pip3 install tensorflow - Installing Keras -
pip3 install keras
Other libraries used:
- Numpy -
pip3 install numpy - base64 -
pip3 install base64- to encode and decode images to and from base64 string - PIL -
pip3 install image- to support image operations like open, save, resize, etc. - imageio -
pip3 install imageio- easy way to get 2d numpy array from an image
Start off by installing create-react-app module
npm install -g create-react-app
Run this module to install default application build
create-react-app my-app
After installation finishes, follow instructions on the console.
For more details, check official tutorial
- react-file-base64 -
npm install react-file-base64 -- save- base64 component - react-drawable-canvas -
npm install react-drawable-canvas --save- ready to use canvas component