indupriyakompi / mnist-kannadadigits

MNIST-KannadaDigits

As a part of this project,I tried my hands in writing up a small book/article to understand the fundamentals of Convolutional Neural Network. The end goal of this book is to make illustrate CNN as easily readble, understandable and learnable to those who're new to Deep Learning. This repository has two main files namely,

1. Bantaba project - Introduction to Convolutional Neural Network using TensorFlow.pdf

This files is pdf file that contanis the basics concepts of CNN usign TensorFlow and Keras, it serves as a guide to the beginners handbook. The mathematical concepts are animated with images for easy understanding.

2. Bantaba Project on Convolutional Neural Network with TensorFlow.ipynb

A very easy way to explain and interpret a CNN is using an image classification model. In this project, we will be learning how CNN classifies Kannada digits, from 1 through 9. This is a 5 layers Sequential Convolutional Neural Network for digits recognition trained on Kannada digits dataset. The sole purpose for having chosen this data set is because it is easier to understand the operation of CNNs through MNIST datasets. I have chosen to build it with keras API (Tensorflow backend) which is very intuitive. To ensure the model did not overfit, we used Keras callbacks.