Transform Day time to night time images for Autonomous Drive Using CycleGAN

This project is to generate night time equivalent images from daytime images, such that had access to some outdoor images acquired at night and some at day, but no pairwise images are available.

CycleGAN model architecture used as it comprised of two generator models and each generator has a corresponding discriminator model.

The discriminator and generator models are trained in an adversarial zero-sum process, like normal GAN models.

The project goal is to train such a 2 set Generator-Discriminator pair, to finally generate night-time images from day-time ones and quantify performance.