Catch up on stories from the past week (and beyond) at the Slashdot story archive

 



Forgot your password?
typodupeerror
×
Anime Japan Software Technology

Japan Researchers Develop Automated Technique For Anime Colorization Using Deep Learning (brightsurf.com) 77

Japanese researchers have developed a technique for automatic colorization in anime production. "To promote efficiency and automation in anime production, the research team focused on the possibility of automating the colorization of trace images in the finishing process of anime production," reports Brightsurf. "By integrating the anime production technology and know-how of IMAGICA GROUP and OLM Digital with the machine learning, computer graphics and vision technology of NAIST, the research team succeeded in developing the world's first technique for automatic colorization of Japanese anime production." From the report: After the trace image cleaning in a pre-processing step, automatic colorization is performed according to the color script of the character using a deep learning-based image segmentation algorithm. The colorization result is refined in a post-process step using voting techniques for each closed region. This technique will be presented at SIGGRAPH ASIA 2018, an international conference on computer graphics and interactive techniques, to be held in Tokyo, Japan on Dec. 4-7. While this technique is still in the preliminary research stage, the research team will further improve its accuracy and validate it in production within the anime production studio. The product of this will be available for commercialization from 2020.
This discussion has been archived. No new comments can be posted.

Japan Researchers Develop Automated Technique For Anime Colorization Using Deep Learning

Comments Filter:
  • How many frames (Score:5, Interesting)

    by AHuxley ( 892839 ) on Thursday November 29, 2018 @05:04AM (#57719500) Journal
    How many artists will get to start their own work rather than filling in frames for others?
    • Re:How many frames (Score:5, Insightful)

      by rhsanborn ( 773855 ) on Thursday November 29, 2018 @07:19AM (#57719694)
      How many artists won't get the entry-level apprentice-like opportunity to learn from people with more experience because this role has been automated?
      • Re:How many frames (Score:5, Informative)

        by mikael ( 484 ) on Thursday November 29, 2018 @08:03AM (#57719778)

        Disney offshored that work from Hollywood to the Far East decades ago, then brought it back when it was automated by computer. Artists from the art schools are way more qualified than to inbetween and fill in paint cells.

        If you look at how they did many of those early animations, there were motion artists who took movies of dancers, rotoscoped the animation to get generic stick figures for reference. These frames were a closely guarded secret as they were reused between animations. When it came to make an animation, they took those reference frames, outlined each with a simple body, get detail artists do a fully detailed frame, then hand those over to inbetweeners who did the intermediate animation frames, and finally to the cel painters who outlined and filled in each individual frame. The deep learning system replaces these stages.

        https://en.wikipedia.org/wiki/... [wikipedia.org]

      • This is the general problem across the first world economies. The gap between entry level and skilled level employment.
        The good old traditional work your way up method we hear.
        0. Get a degree (lets say accounting)
        1. Start a job in the mail room, you get to know the peoples name and the departments.
        2. Move up to delivery where you get to know the people.
        3. You find begin to interact with the accounting department
        5. Do some internship with that department, enough to show that you know your stuff.
        6. Get into A

      • How many artists won't get the entry-level apprentice-like opportunity to learn from people with more experience because this role has been automated?

        Note that much anime grunt work in Japan has been farmed out to Korea where the labor is cheaper. Just look at the credits of any anime that aired in the last several years -- almost all Korean names. Not only the coloring but the in-betweeners.

        So the problem you are pointing out already happened. Some jobs will remain but mostly for the higher-end stuff.

  • Pass (Score:4, Interesting)

    by stealth_finger ( 1809752 ) on Thursday November 29, 2018 @05:05AM (#57719504)
    Once you start to get 3d assets and computerised stuff in anime it loses that look which gives it that x factor. It's too clean and looks wrong. IMO at least.
    • by GeLeTo ( 527660 )
      This is done with Deep Learning. If the source material used to train the neural network has the "x factor" and does not look clean - so should the output result.
      • This is done with Deep Learning. If the source material used to train the neural network has the "x factor" and does not look clean - so should the output result.

        By its own operation it's going to smooth everything out when it's trying to make an amalgamation of all its source material. It'll probably look good sure but will be missing something.

        • By its own operation it's going to smooth everything out when it's trying to make an amalgamation of all its source material.

          Deep Learning is good at representing complex patterns, such as which designs can or cannot be used together, so that's most likely not going to be a problem. A standard technique is to use adversarial networks to provide feedback on the quality by comparing the generated material to original sources. Anything that stands out, such as excessive smoothing, will be used to correct the generating network.

        • by AmiMoJo ( 196126 )

          The biggest reason for the declining quality is the reduced time given to production. Anything that speeds it up and leaves more time for review should improve the animation.

    • I agree. Just look at all the awful netflix ones.
      • I disagree. I actually quite like the quality of the artwork in the Netflix ones. The big problem with the Netflix ones is that more often than not the story is absolute garbage.

        • by zlives ( 2009072 )

          i for one like
          B: The Beginning
          Knights of Sidonia
          castlevania
          and others
          not sure if they are all netflix but i for one enjoy them

    • South Parks anime rip looked pretty good.
    • by AmiMoJo ( 196126 )

      It's been happening slowly for a couple of decades now. An early example would be Azumanga Daioh, which was drawn in pencil as usual but then scanned and inked and coloured digitally. It looks pretty normal, most people at the time didn't realize.

      Later even Studio Ghibli started using digital stuff for in-between frames, although all the key stuff is hand painted.

      3D and digital has massively reduced the cost of producing anime and vastly improved the quality. Early 3D stuff did stand out, but now when they

      • I was watching a video on youtube the other day with some guy trying to address it, because hand drawing that stuff takes stupid amounts of time. Basically he 3d modelled this ship, animated it and exported it as lines and coloured and animated over the top. It did actually look pretty good at the end.
    • by mikael ( 484 )

      I agree. Ulysses 31 is a good example. The French did the story lines, but they needed the Japanes Anime artists help to really give those animations a means to convey intense explosions. Sometimes they did that by mixing purples blues and whites, for supernova, red/orange/yellow/black for an exploding planet, or black/white cracks with god rays for strange exotic matter.

    • by Anonymous Coward

      How much modern Anime is CGI. You can only tell when they fall down (like making a character too detailed so the cell shading ends up with artifacts instead of large flat segments that don't change much during a scene.)

      Everything I have watched this year is CGI, and most either uses 'grain' over the CGI, or just leaves it in. Especially bad are the Americanized CGI Animes, like new Berserk, Fist of the North Star Regenesis, Baki (Netflix's adult oriented stab at these), Goblin Slayer (which I can't recommen

    • Once you start to get 3d assets and computerised stuff in anime it loses that look which gives it that x factor.

      Anime loses it's great and compelling stories, character development, excellent voice acting, and overall joy to watch because a computer was involved in drawing?

  • Text article (Score:5, Insightful)

    by religionofpeas ( 4511805 ) on Thursday November 29, 2018 @05:31AM (#57719532)

    What's the point of a short text article about graphics without adding any pictures/video clip as demonstration ?

    • Mostly likely this deep learning algorithm found the GWBASIC PAINT command.

      10 SCREEN 7 ' It is 2018 lets use EGA Graphics
      20 BLOAD "FRAME1.BLD"
      30 PAINT 160, 100, 1

      chances are I just filled in something blue!

    • From the reference in the article to submitting the paper at SIGGRAPH Asia, this announcement is an attempt to drum up interest in advance of the conference; releasing multimedia prior to the conference would dilute the interest -- "oh, I've already seen that; I'll go watch a _new_ presentation" -- and reduce the attendance at their presentation. Expect it to be flogged much more heavily, with extensive stills and video, after the conference.
  • To whomever wrote this: thank you for calling this Deep Learning instead of going "OMG AI".
  • Interesting how it will perform on it :)
  • Too bad anime plots seem computer generated these days.

  • In related news, there's a program called "waifu2x" that uses deep learning and neural networks (and CUDA, if you have an NVIDIA graphics card...highly recommended, there's a ~4x speed boost on my machine) to automatically scale up ("super-resolution") and de-noise anime art: https://github.com/nagadomi/wa... [github.com]

    For Windows users, use this version: https://github.com/lltcggie/wa... [github.com]
  • RTFA and there's nothing there.... just "we can do this" but no pics (likely because they're going to announce the technique formally in Dec). I expect that this will look more or less something like the cell-shaded early CGI animation.

    Obviously, this is about "optimizing" the drudgework of colorizing, meaning the teams of sub-artists on productions will just be a little smaller.

    I guess everything depends on how persuasive this looks; it could be cool and seems like a relatively low-hanging fruit for autom

  • by Anonymous Coward

    I bet we will see some weird mistake, like pink hair.

  • Comment removed based on user account deletion

Single tasking: Just Say No.

Working...