Hallucination from noon to night images using CNN

Given that the deep neural networks (DNNs) can now achieve almost human level performance in the object classification task, a question arises whether DNNs can execute meaningful color transforms of images. The specific goal of our color transfer is to imitate time transfer from noon to night on an image. We eliminate the demand for training pairs of noon and night images by adopting adversarial network [Goodfellow et al. 2014].