A GAN Based Automatic Image Colourizer

Samridh  Agarwal; Vidoosh Kumar  Bang; Parthvi  Singh; B.K.  Tripathy

Abstract

A fascinating skill in computer vision under Artificial intelligence is the technique of automatically changing monochrome photos into colour ones without human assistance. When a substantial annotated dataset is not readily accessible, transfer learning is commonly employed as the preferred approach. Typically, this involves utilizing pre-trained models with extensive general knowledge gained from training on ImageNet, which are subsequently fine-tuned to cater to specific tasks. From the utility point of view, it is useful in the restoration of movies to the augmentation of photos to increase their readability, re-colorization of pictures, improvement of grayscale drawings, and the restoration of black-and- white photographs. It is believed that automated colorization, in the case of black-and-white movies, has the potential to improve the viewing experiences of viewers. It is possible to utilize it in the sphere of security as well since it can transform the black-and-white photos collected by CCTV cameras into colour ones. In this study, we suggest a method for colouring grayscale images using a Generative Adversarial Network (GAN).

Abstract

Keywords