Black and White Image Colorization using Deep Learning

The computerized colorization of black and white images began in the 1970's., after which image colorization has come a long way. RGB images are widely used in many fields to get additional information regarding the image that cannot be obtained from B&W images. Some areas where RGB images might be used in determining the chemical composition of a material, chromatographic spectroscopy and studying the biology of plants. In this system we approach the image colorization problem using deep learning techniques. We achieve this using Conditional Generative Adversarial Networks (cGAN) which will be trained on the COCO dataset. This is a fully automated model and requires no human assistance to achieve artifact-free quality images. The recent achievements in Generative models have inspired this system. We present two models in this system, of which the second model produces better results with only one-fourth of the training time of the first model.