Stories
Slash Boxes
Comments

SoylentNews is people

posted by chromas on Tuesday March 19 2019, @10:23PM   Printer-friendly

Nvidia AI turns sketches into photorealistic landscapes in seconds

Today at Nvidia GTC 2019, the company unveiled a stunning image creator. Using generative adversarial networks, users of the software are with just a few clicks able to sketch images that are nearly photorealistic. The software will instantly turn a couple of lines into a gorgeous mountaintop sunset. This is MS Paint for the AI age.

Called GauGAN, the software is just a demonstration of what's possible with Nvidia's neural network platforms. It's designed to compile an image how a human would paint, with the goal being to take a sketch and turn it into a photorealistic photo in seconds. In an early demo, it seems to work as advertised.

YouTube video (1m59s).


Original Submission

This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
(1)
  • (Score: 5, Funny) by c0lo on Tuesday March 19 2019, @10:34PM (3 children)

    by c0lo (156) Subscriber Badge on Tuesday March 19 2019, @10:34PM (#817148) Journal

    Game studios can now let go another wave of artists.

    --
    https://www.youtube.com/watch?v=aoFiw2jMy-0 https://soylentnews.org/~MichaelDavidCrawford
    • (Score: 3, Disagree) by takyon on Tuesday March 19 2019, @11:41PM (2 children)

      by takyon (881) <takyonNO@SPAMsoylentnews.org> on Tuesday March 19 2019, @11:41PM (#817171) Journal

      2D landscapes don't make a game.

      In general though, machine learning algorithms could help to make larger, more populated, and more complicated game worlds, or lay off game designers. Or both.

      --
      [SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
      • (Score: 5, Interesting) by Anonymous Coward on Wednesday March 20 2019, @12:21AM

        by Anonymous Coward on Wednesday March 20 2019, @12:21AM (#817178)

        2d makes is for skyboxes, for far away things. And he said another wave, this time those doing matte painting or even concept artists, not all of the artists. Just draw some zones, and bam! you got a quick concept for a space city with thing filled by the machine learning (artificial dumbness, not intelligence). It's just a set of waves, we can make a poll about what will be the next one. Maybe next they will try to replace basic 3D asset creation (textures and shapes).

        Personally I think they suck in some cases, with crappy halos and weird textures. But that will not mean cuts will not be made soon, 4 overworked painters instead of 5, trying to fix the results of automatic tools supposed to help. Just like the rushed "photoshops" seen in magazines in last 10-20 years, with people having the wrong number of limbs because someone got a bit happy with the clone tools and nobody had no time to check. Profit over quality, and remember, time is money.

        BTW, the demo video shows screen with Ubuntu?

        Their biggest challenge is how to automatize customers, because so far the proles are getting pennyless with so much firing. AC, Artificial Consumer, fully automatized economy... do humans dream of automatic nightmares?

      • (Score: 3, Informative) by c0lo on Wednesday March 20 2019, @02:39AM

        by c0lo (156) Subscriber Badge on Wednesday March 20 2019, @02:39AM (#817212) Journal

        2D landscapes don't make a game.

        On top of the insightful notes in the comment above [soylentnews.org] (including the "automatic nightmare" one), consider these:
        - Google’s DeepMind develops AI that can render 3D objects from 2D pictures [venturebeat.com] (static case)
        - A New AI-Powered System Creates Impressive 3D Animations From Still Images [gizmodo.com.au] (3D extraction and rigging)

        --
        https://www.youtube.com/watch?v=aoFiw2jMy-0 https://soylentnews.org/~MichaelDavidCrawford
  • (Score: 1) by ensigndna on Wednesday March 20 2019, @12:25AM (1 child)

    by ensigndna (7179) on Wednesday March 20 2019, @12:25AM (#817181)

    Holy cum guzzling gutter sluts, Batman.... this sure is neat!

    • (Score: 2) by DannyB on Wednesday March 20 2019, @02:05PM

      by DannyB (5839) Subscriber Badge on Wednesday March 20 2019, @02:05PM (#817330) Journal

      Your comment brings to mind a potential application for this technology.

      Looking at the UI in TFA, I would imagine you replace the buttons: Trees, Rock, River, Sky, etc, with new buttons having imaginative titles like Press Secretary, Chief of Staff, White House Counsel, etc.

      --
      To transfer files: right-click on file, pick Copy. Unplug mouse, plug mouse into other computer. Right-click, paste.
  • (Score: 0) by Anonymous Coward on Wednesday March 20 2019, @01:21AM (1 child)

    by Anonymous Coward on Wednesday March 20 2019, @01:21AM (#817189)

    What's next, creating photorealistic images of people from sketches? Police sketch artists are going to have some fun with that.

    • (Score: 2) by legont on Wednesday March 20 2019, @10:14AM

      by legont (4179) on Wednesday March 20 2019, @10:14AM (#817289)

      Yeah, photo-realistic pussy galore.

      --
      "Wealth is the relentless enemy of understanding" - John Kenneth Galbraith.
  • (Score: 0) by Anonymous Coward on Thursday March 21 2019, @12:49AM

    by Anonymous Coward on Thursday March 21 2019, @12:49AM (#817641)

    Wow, this could be fun. I see the are running it on Ubuntu. I want copy for local install, please. Pretty please. 8)
    I would feed it all my images and then have a load of fun creating arty landscapes.

(1)