JamesQuirk/CNN-From-Scratch

CNN From Scratch

This 'CNN From Scratch' is a simple CNN framework that has been coded using only NumPy for matrix mathematics. This project was done to explore, in detail, how Convolutional Neural Networks work.

Diagram

The diagram below shows an example forwards flow of a CNN.

Useful References

• TowardsDataScience - Convolutional Neural Networks from the ground up

• TowardsDataScience - Applied Deep Learning

• Convolutional Networks

• MNIST database

• SuperDataScience - The Ultimate Guide to Convolutional Neural Networks (CNN)

• Convolutions and Backpropagations

• 3Blue1Brown Backpropagation calculus

• ML Cheatsheet

• Medium - Backpropagation for Convolution with Strides [> 1, w.r.t. inputs]

• Medium - Backpropagation for Convolution with Strides [> 1, w.r.t. filters]

• On Vectorization of Deep Convolutional Neural Networks for Vision Tasks