I remember one episode in Star-Trek The Next Generation where they had a picture of some guy with more than half of his face covered by someone else.
The computer restored the full face by reconstructing it in "3D".
Sounded like treknobabble gibberish back in 1994, but it seems that we're getting closer and closer to such abilities.
Our brains can recognize things by seeing only (small) parts of them, filling the rest with imagination, why wouldn't computers be able to do the same ?
One step further to Strong AI.
That was my thought too. I was only half paying attention to the video, I saw that and had to double take. I'm skeptical about that. Crazy how good this if that's true. Even the less extreme ones are crazy.
Hehe, I guess you're not ready for this, but, your kids are going to love it. xD
P.S. I'd like them to take it a step forward with additional detail via a.i. , like take an old mobile phone picture that's e.g. 640x480 or 1024x768 and boost the detail up to full hd and beyond.
The question is when do we get to use this? Any plugins for Photoshop?
AI plugin? Like 2GB plugin? It is likely same platform as one used for restoration of partially deleted images as it quite well "restored" image which was missing information due to Text. It uses object search to find expected object shape/structure. Actual denoiser which was used in that progressive example could be reasonably small. But entire package which did all those images...
I reckon you've heard about the latest addition to Top500: Sierra (#1) and Summit (#3)
(congratz USA, gz Nvidia and IBM)
Wanna hear the freaky fact about them?
Taken together, Sierra, Summit and previous #1 (current #2, Sunway TaihuLight)), taken together these three supercomputers represent more deep learning capability than the other 497 systems on the TOP500 list combined,
AI plugin? Like 2GB plugin? It is likely same platform as one used for restoration of partially deleted images as it quite well "restored" image which was missing information due to Text. It uses object search to find expected object shape/structure. Actual denoiser which was used in that progressive example could be reasonably small. But entire package which did all those images...
With A.I. behind it, then, there's possibly 2 ways; 1) Traditional add-on, or 2) Ability to use the internet as an additional source.
Camera manufacturers dream. Get less noise with high ISO. The denoiser in LR mostly blurs the image - bad. Color denoiser works good, but makes the colors looks a tad bit muddy. I can see this being used in every camera and image editing softwares IF nVidia is ready to sell it at a reasonable price.
This can be particularly useful in phones, because of their tiny sensors which are extremely noisy.
I can imagine AI be used more extensively in gaming, ie, facial expressions, animal and plant movements, environments, etc. And probably replace traditional AA sampling with AI implementations with much less of a performance hit.
I can imagine AI be used more extensively in gaming, ie, facial expressions, animal and plant movements, environments, etc. And probably replace traditional AA sampling with AI implementations with much less of a performance hit.
It's probably gonna be the same deal as PhysX or GameWorks. Meaning an in-game option that should be labelled "30FPS limiter" :P
R.i.p watermarks on pictures and graphs.;)