Deep Koalarization: Image Colorization using CNNs and Inception-ResNet-v2

TL;DR

This paper introduces a deep learning model combining CNNs and Inception-ResNet-v2 features for image colorization, capable of handling various image sizes and validated through user studies and diverse applications.

Contribution

It presents a novel encoder-decoder architecture that integrates high-level features from Inception-ResNet-v2 for improved image colorization from grayscale images.

Findings

The model successfully colorizes images of any size and aspect ratio.

User studies indicate high acceptance of the colorized images.

Application examples include historical photographs and diverse image types.

Abstract

We review some of the most recent approaches to colorize gray-scale images using deep learning methods. Inspired by these, we propose a model which combines a deep Convolutional Neural Network trained from scratch with high-level features extracted from the Inception-ResNet-v2 pre-trained model. Thanks to its fully convolutional architecture, our encoder-decoder model can process images of any size and aspect ratio. Other than presenting the training results, we assess the "public acceptance" of the generated images by means of a user study. Finally, we present a carousel of applications on different types of images, such as historical photographs.