A group research project in associate with Boming Shi.
Generative adversarial network(GAN) has achieved great progress in many applications. Also the combination of ”top-down” discriminator and ”bottom-up” generator has provided a new thought to generative model. In the paper ”Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks”, Zhu et al. presented cycleGAN to implement unpaired image translation. Although the results shown in the paper are impressing, there are some drawbacks in cycleGAN, such as mode collapse and difficulty in transfiguration. In our paper, our goal is to explore the performance of cycle consistency loss and improve cycleGAN with Wasserstein GAN.