The longest discussion about this article on hacker news so far: https://news.ycombinator.com/item?id=2578706
I wonder why Microsoft payed him to do develop this kind of algorithm.
I wonder why Microsoft payed him to do develop this kind of algorithm.
Very curious to see this as it progresses however.
It's not perfect but seeing the alternatives it is a huge step in the right direction.
a a matter of fact some of the most recents algo available for video deinterlacing like nnedi3 are perfect for that, and of course also work for pixel art, and I think they blow away the example showed in their gallery - https://www.youtube.com/watch?v=0691zsXWbhA
I wonder if now, in the tensorflow era, a better automatic approach using generative nn could be devised to reach beyond 8x