Landscape colorization using a classification approach

Context

This project is developed in the context of the lecture at the TU Wien, the goal was to find a challenging topic and build a Deep Learning model to solve it.

The project

Colorizing grayscale images is an old topic in machine learning but we’ve all seen the greyish tasteless colorized images system usually output (see aside).

This is usually due to the regression approach system employs. In this project, I approached the problem differently and categorised the colors. With such an approach, the results are more colourful, the models are not tempted to predict grey-brown colors.

Example of image colorized by a classic colorizer

The following video briefly describes the problem, the approach and and results I obtained. You can also have a look to the accompanying report here.