Hacker News new | ask | show | jobs
by kore 3370 days ago
This has also been making the rounds, Deep Photo Style Transfer: https://github.com/luanfujun/deep-photo-styletransfer
3 comments

I inquired about the license for that one but unfortunately they can't allow commercial use.

https://github.com/luanfujun/deep-photo-styletransfer/issues...

These are amazing.

I wonder if it could start with input of a normal portrait of you, style: some celebrity of your gender, and output: what you would look like after receiving the same style? It doesn't seem beyond the examples shown at your link...

It would bring a whole new meaning to the word "filter" (instagram, etc.) I particularly like that the original is very much present in the output: it would still be "you".

But maybe there are subtle problems that I don't notice because I don't look at the subject matter as carefully. People pay a lot of attention to recognizing each other, perhaps the effect would not transfer as well as these examples presented.

Unfortunately, no one seems to be able to run it.
This person ran it on a hackintosh (I helped a little) https://github.com/luanfujun/deep-photo-styletransfer/issues...
Would appreciate a link showing this.
Believe you'll find supporting comments on the original HN thread: https://news.ycombinator.com/item?id=13958366

Matlab seems to need a commercial license for at least one barrier.