Recognizing Digits using TensorFlow.js in Google Chrome

In this blog post, we will create a simple web application that provides a canvas (mobile desktop laptop tablet ready) for the user to draw a digit and uses a deep neural network (MLP or CNN) to predict what digit the user had drawn. As we already know the capabilities offered by TensorFlow.js, we will extend the ideas to create two Deep Neural Networks (MLP and CNN) in Keras Python environment to recognize digits and use TensorFlow.js to predict the user drawn digit on a canvas in a web browser. In this learning path, we will restrict the user to draw a single digit between [0, 9] and later we will extend the idea to multiple digits. Below is the interactive demo that you can use to draw a digit between [0, 9] and the Deep Neural Network (MLP or CNN) that is running in your browser will predict what that digit is in the form of a bar chart. Important: This is a highly experimental demo for mobile and tablet devices.