Notes

Automated colorization of black and white images has been subject to much research within the computer vision and machine learning communities. Beyond simply being fascinating from an aesthetics and artificial intelligence perspective, such capability has broad practical applications ranging from video restoration to image enhancement for improved interpretability.
We have used Self – Attention Generative Adversial Networks as a deep learning model and modified it to produce realistic and authentic looking results.
Normally, even creating a Deep learning model for colorizing old images using a traditional CNN, the results were quite dull and not authentic looking. In contrast, GANs effectively replace those hand coded loss function with a network – the critic/discriminator – that learns all this stuff for you, and learns it well. Hence, we used GANs to solve the problem of realistic colorization.

1.2 Background and Motivation
Today, colorization is done by hand in Photoshop. To appreciate all the hard work behind this process, take a peek at this gorgeous colorization memory lane video. In short, a picture can take up to one month to colorize. It requires extensive research. A face alone needs up to 20 layers of pink, green and blue shades to get it just right.
Neural nets are fabulous in dealing with bad/incomplete data, which is why color can be added successfully even if the photos are in bad shape. Note that I say “believable coloring” because colorization is an “unconstrained” problem. That is to say, there’s no one right color for a lot of things (like clothes, for example).
Restoration, for the purposes of this project, would be taking the further step of trying to believably replace details where they’re missing or damaged. To me, the most common problem I see with old photos is that they’re faded, so my first ambition is to have another neural network “undo” that fade.
1.3 Objectives
Colorizing and restoring an old image with the help of a GAN. It involves:
• Colorizing Black and white images.
• Replacing missing or damaged details.
• Reducing Fadedness in old images.