Skip to content

Latest commit

 

History

History
27 lines (18 loc) · 1.14 KB

README.md

File metadata and controls

27 lines (18 loc) · 1.14 KB

Image Colorization

Automatic image colourization is the task of introducing colour to a black and white image in a realistic fashion, without any aid from the user. This has long been the focus of research, partly due to its possible applications in image processing and animation.

Recently, deep learning methods, specifically convolutional neural networks, have shown great promise in image processing tasks, as image datasets become more representative and computational power becomes more ubiquitously available. In this work, I focus on the application of these networks to the problem of image colourization, implementing three very different deep neural networks.

Contents

This repository contains several implementations of methods to preform colorization:

1. A deep CNN with a MSE loss

2. A deep CNN using categorical cross entropy loss (Based on http://richzhang.github.io/colorization/)

3. A GAN

Results

demo

demo