Image-to-Image Translation

Also see Sim2Real.

This is really neat:

I am not sure if I can really put domain adaptation and Image-to-Image Translation in the same umbrella.

They seem to be done using GANs, however in the past 2 years Diffusion Models have drastically outperformed them. This is actually sooo interesting.

Image-to-image translation is super important.

From this paper: (Unsupervised Image-to-Image Translation)

Many computer visions problems can be posed as an image-to-image translation problem, mapping an image in one domain to a corresponding image in another domain. 
- super-resolution = mapping a low-resolution to a high-resolution image
- colorization = mapping a gray-scale image to a corresponding color image
- mapping a night image to a day image