The particularly cynical among us have always said to believe half of what you see and none of what you hear, but it may be time for even the most trusting to adjust that ratio somewhat. I feel this way because I’ve just seen some incredible examples of what (I think) is called Unpaired Image-to-Image Translation, and I assume a catchier name is forthcoming.
Unpaired Image-to-Image Translation uses “Cycle-Consistent Adversarial Networks” to transform images in ways you might not have thought possible. It can make paintings look like photos. It can make photos look like paintings. It can make a photo of a bunch of apples look like a photo of a bunch of oranges. It can change horses to zebras. And so on and so forth.
You can get a very technical breakdown of Unpaired Image-to-Image Translation right here, and see a few examples in the gallery below. Plus side: You now have a plausible defense if anyone ever gets a photo of you doing something embarrassing.
Here’s the gallery: