Style transfer is an amusing field, and a lot of parallel research is going on across different research groups to improve the state of the art. This chapter has discussed the two most influential works in the paired and unpaired style transfer space so far. There have been a few more related works in this space that are worth discussing.
There are several other works in the same space. For the sake of completeness and consistency, we limit our discussion to only a few of them. You are encouraged to explore other interesting architectures as well.
DiscoGAN
Kim and Cha et al. presented a model called DiscoGAN that discovers cross-domain relations with GANs. Transforming black-and-white images into colored images, satellite images into map-like images, and so on can also be termed cross-domain transfer and style transfer.
As we’ve seen, cross-domain transfer has several applications in the real world. Domains such as autonomous driving and healthcare have started to leverage deep learning techniques, yet many use cases fall short because larger datasets are unavailable. Unpaired cross-domain transfer works such as DiscoGAN (and CycleGAN) can be of great help in such domains.
Get hands-on with 1400+ tech skills courses.