Recommendation System Recommending Natural Colours for Black and White Images

Abstract

In the past ten years, the idea of automatic image colorization has attracted attention for a range of uses, including the restoration of old or damaged photos. This problem is extremely poorly presented since assigning colour information involves such a wide range of degrees of freedom. Recent developments in automatic colorization frequently use input images that share a common theme or data that has undergone extensive processing, like semantic maps. Using conditional adversarial networks, we attempt to fully broaden the colorization process and address image colorization issues. Landscape colour and grayscale images from the publicly accessible Kaggle dataset were used to train the network.