About this fork

This fork is a fork of vzhou842/cnn-from-scratch with the following changes:

• implemented multi-channel input for Conv3x3
• refined the computation of backward propagation in Conv3x3
• provided a simple example of building two layers of Conv3x3

The original README.md is as follows.

A Convolution Neural Network (CNN) From Scratch

This was written for my 2-part blog post series on CNNs:

• CNNs, Part 1: An Introduction to Convolution Neural Networks
• CNNs, Part 2: Training a Convolutional Neural Network

To see the code (forward-phase only) referenced in Part 1, visit the forward-only branch.

Usage

Install dependencies:

$ pip install -r requirements.txt

Then, run it with no arguments:

$ python cnn.py
$ python cnn_keras.py

You can also run this code in your browser.

More

You may also be interested in a Neural Network implemented from scratch in Python, which was written for my introduction to Neural Networks.