❤️ Train a neural network and track your experiments with Weights & Biases here:
📝 The paper “InfiniteNature-Zero Learning Perpetual View Generation of Natural Scenes from Single Images” is available here:
❤️ Watch these videos in early access on our Patreon page or join us here on YouTube:
–
–
🙏 We would like to thank our generous Patreon supporters who make Two Minute Papers possible:
Aleksandr Mashrabov, Alex Balfanz, Alex Haro, Andrew Melnychuk, Benji Rabhan, Bryan Learn, B Shang, Christian Ahlin, Eric Martel, Geronimo Moralez, Gordon Child, Ivo Galic, Jace O’Brien, Jack Lukic, John Le, Jonas, Jonathan, Kenneth Davis, Klaus Busse, Kyle Davis, Lorin Atzberger, Lukas Biewald, Luke Dominique Warner, Matthew Allen Fisher, Michael Albrecht, Michael Tedder, Nevin Spoljaric, Nikhil Velpanur, Owen Campbell-Moore, Owen Skarpness, Rajarshi Nigam, Ramsey Elbasheer, Steef, Taras Bobrovytsky, Ted Johnson, Thomas Krcmar, Timothy Sum Hon Mun, Torsten Reil, Tybie Fitzhugh, Ueli Gallizzi.
If you wish to appear here or pick up other perks, click here:
Thumbnail background design: Felícia Zsolnai-Fehér –
Károly Zsolnai-Fehér’s links:
Instagram:
Twitter:
Web:
Comments
30 responses to “Google’s New AI: Fly INTO Photos…But Deeper! 🐦”
I freaking love seeing progress in technology like this. Always happy to see a new Two Minute Papers video!
What a time to be alive!
Maybe this technique can be combined with real data from Google Earth in the future, so the generated images would depict the real terrain where that photo was made.
imagine future open world games procedurally generated from a single mood setting image
But can you fly back to where you have come from and see the same landscape? Only if it’s saving the past. right?
its horrible
can we test it ??
Can you please stop calling me a fellow scholar? I'm not a scholar. Thank you.
Dude, image flying around in any fictional world from a single picture. So awesome!
The shifting effect reminds me of my dreams where the scenery and objectives and story of my dreams are constantly shifting and changing and as it goes on, one minute I can be in one place doing one thing and the next minute something entirely different is happening in an entirely different place, like literally I turn around in the dream and am somewhere else completely but somehow the impossible transitions make sense.
It seems like the processes the ai goes through to create new images as it attempts to dig deeper into a picture is similar to the process our brain goes through as it attempts to piece together a coherent narrative while we're dreaming.
This is just incredible and so exciting as a digital artist 🤯
Damn all this AI technology is evolving at such a fast rate, I'm finding it hard to keep up.
it's actually super boring. and it's all on a screen. who cares?
This technic requires two opposite live video feed to work properly, AI algorithms should easily fulfill the rest, with something like this virtual tourism would become a good business model
Wait so it just makes it up as it goes? So it’s imagining things?
Google found the source code to our simulation and is making a simulation within the simulation
Any two minutes paper alternative channel?
looks like an acid trip
holding on to my papers….
soon google ai will be able to fly into my bedroom
This is how I imagine 4D looking like
GTA in Google Street, when?
What a time to be alive! 🎉
Somehow, this looks like fractals when the camera moves faster
This is so dreamlike for me. The way things morph and blend and it all makes sense but it also doesn't… like a dream.
It's like turning the real world into a fractal. Very cool. Also something I have done in my head before, the 'zooming into' a visualization. I was doing it the other night while falling asleep, usually leads you right into a vivid dream.
I'm glad none of these videos are two minutes anymore
how???
This really looks like some parts of my dreams, but in dreams it works with attention and level of detail more. The more you pay attention to something, the more detail you get, and if I do not pay attention much, it looks something like this, but a bit more stable across time. (or maybe because I don't move so fast)
Last time I tried examining shadow simulation in my dreams, it was pretty impressive, not realistic at all, but quite complex. But when I tried if my brain can properly cast shadows of my hand and my 5 fingers, it could not. 😀 It clearly knew that it is supposed to do something more/smart , but my brain could not intuitively make it happen. I think because the subconscious is bad at counting and logic. 😀
If this technique continues to improve, creating more accurate detail with every generation, Will this technique eventually be able to re-create me down to the individual particles in every atom in my body?
there is an text to video generator from Meta AI