Creating Virtual Worlds 20x Faster!
6:07

Creating Virtual Worlds 20x Faster!

Two Minute Papers 05.11.2023 97 435 просмотров 4 794 лайков

Machine-readable: Markdown · JSON API · Site index

Поделиться Telegram VK Бот
Транскрипт Скачать .md
Анализ с AI
Описание видео
❤️ Check out Weights & Biases and sign up for a free demo here: https://wandb.me/papers 📝 The paper "3D Gaussian Splatting for Real-Time Radiance Field Rendering" is available here: https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/ Luma Labs: https://twitter.com/nickfloats/status/1709323018852319522 In Unreal Engine: https://www.unrealengine.com/marketplace/en-US/product/3d-gaussians-plugin Text to 3D: https://gsgen3d.github.io/ + paper https://arxiv.org/abs/2309.16585 Dream Gaussian (image to 3D): https://dreamgaussian.github.io/ 25 photos: https://twitter.com/AlexandreDevaux/status/1707365670596284857 Animation: https://twitter.com/simesgreen/status/1710691657740341428 My latest paper on simulations that look almost like reality is available for free here: https://rdcu.be/cWPfD Or this is the orig. Nature Physics link with clickable citations: https://www.nature.com/articles/s41567-022-01788-5 🙏 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, Bret Brizzee, Bryan Learn, B Shang, Christian Ahlin, Gaston Ingaramo, Geronimo Moralez, Gordon Child, Jace O'Brien, Jack Lukic, John Le, Kenneth Davis, Klaus Busse, Kyle Davis, Lukas Biewald, Martin, Matthew Valle, Michael Albrecht, Michael Tedder, Nikhil Velpanur, Owen Campbell-Moore, Owen Skarpness, Rajarshi Nigam, Ramsey Elbasheer, Richard Sundvall, 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: https://www.patreon.com/TwoMinutePapers Thumbnail background design: Felícia Zsolnai-Fehér - http://felicia.hu Károly Zsolnai-Fehér's research works: https://cg.tuwien.ac.at/~zsolnai/

Оглавление (4 сегментов)

Introduction

Because so many of you Fellow Scholars have  requested it, we had a video on one of the   best papers in the year in my opinion, which  is a work called Gaussian Splatting. And I am   absolutely speechless by how quickly things are  unfolding around it. I think you will be too. So,   what does this do? Well, just imagine grabbing  your smartphone, taking a few images or videos   of real places, and then, we could create  a virtual version of them and play in them   in real time. And it does exactly that. And  these scenes come out really detailed. Dear Fellow Scholars, this is Two Minute  Papers with Dr. Károly Zsolnai-Fehér. And the key thing here is that this is  a handcrafted technique that is easy   to implement. And now, some publicly available  applications already have it. This one does. Yes,   the dream is coming true, we might just play  in a virtual copy of the real world very soon.    Especially that it is available in Unreal  Engine too through a paid plugin, though   I am sure that free versions will pop up very  soon too. In fact, things are going so quickly,   I wouldn’t be surprised if by the time this  video appears, it is already available for free. And it gets even better. Way better! For instance,  we know that text to image is possible, but we are

Text to 3D

talking about 3D virtual worlds here. So, oh my.   It has been just a few months of this paper being   out, can that really be? Follow up papers are  already popping up. Wait a second. This is text to   3D. We write something, and out comes 3D geometry.   Our last episode was on exactly that, and it is   already being surpassed. What is happening? Wow.   Well, what is happening is that Gaussian Splatting   is already being used to improve these techniques.   Hmm, in many cases, I see more detailed geometry   being produced. And all this in just a couple of  months and one more paper down the line. So cool!

Image to 3D

But it doesn’t end there. Not even  close. Goodness! Here is another   paper which is able to perform  image to 3D. One photo goes in,   and out comes 3D geometry. I already see that it  converges to a solution much quicker than earlier,   neural radiance fields-based solutions,  that is fantastic news, but come on,   what can it do? Wow, it really does it. In  goes the image, and out comes the 3D model. And it gets better. Then, we can even really  put these little models into our virtual worlds,   and animate them. And all this process takes…how  long exactly? The previous paper we looked at took   40 minutes, so what about this one? Now  hold on to your papers Fellow Scholars,   because it is two minutes. This is incredible!   A 20x speedup in just one paper. I really don’t   know what to say here. In the first step, it  lines up the Gaussians, little mathematical   lumps to match the shape of the character  or object, and then, in the second step,   it slowly refines the textures as well. And this  one runs on a single consumer graphics card, this   is what does it in two minutes per object. Haha!   This work is a real match for Two Minute Papers. No wonder these papers caught the imagination of  you Fellow Scholars. It gets even better. It can   even reproduce photorealistic faces reasonably  well with just 25 photos as an input, and…whoa!    Here, you can even animate these inputs. So not  only creating virtual worlds from the real world,   but we can even let our imagination run wild and  create absolutely anything out of it. I love it.

gaussian splatting

Now, just one more thing. I know some  of you Fellow Scholars are asking,   okay but what is Gaussian Splatting? What  does that even mean? Now the Gaussian part   means that objects in this scene will be  represented as a sum of many little waves.    And splatting is a way of throwing these  little waves from the 3D virtual world   at our 2D screen and computing what they  should really look like. And this simple,   but effective idea taking the world by storm.   A handcrafted technique. Powered not by AI, but   by human ingenuity. Once again, the papers coming  alive in the real world. What a time to be alive!

Другие видео автора — Two Minute Papers

Ctrl+V

Экстракт Знаний в Telegram

Экстракты и дистилляты из лучших YouTube-каналов — сразу после публикации.

Подписаться

Дайджест Экстрактов

Лучшие методички за неделю — каждый понедельник