Become a fan of Slashdot on Facebook

 


Forgot your password?
Close
typodupeerror
×
AI Google Communications Network Networking Software Technology

Google Brain Creates Technology That Can Zoom In, Enhance Pixelated Images (softpedia.com) 146

Posted by BeauHD from the zooming-progression dept.
Google Brain has created new software that can create detailed images from tiny, pixelated images. If you've ever tried zooming in on an image, you know that it generally becomes more blurry. You'd just get larger pixels and not a clear image. Google's new software effectively extracts details from a few source pixels to enhance pixelated images. Softpedia reports: For instance, Google Brain presented some 8x8 pixel images which it then turned into some pretty clear photos where you can actually tell facial features apart. What is this sorcery, you ask? Well, it's Google combining two neural networks. The first one, the conditioning network, works to map the 8x8 pixel source image against other high-resolution images. Basically, it downsizes other high-res images to the same 8x8 size and tries to make a match on the features. Then, the second network comes into play, called the prior network. This one uses an implementation of PixelCNN to add realistic, high-res details to that 8x8 source image. If the networks know that one particular pixel could be an eye, when you zoom in, you'll see the shape of an eye there. Or an eyebrow, or a mouth, for instance. The technology was put to the test and it was quite successful against humans. Human observers were shown a high-resolution celebrity face vs. the upscaled image resulted from Google Brain. Ten percent of the time, they were fooled. When it comes to the bedroom images used by Google for the testing, 28 percent of humans were fooled by the computed image.
This discussion has been archived. No new comments can be posted.

Google Brain Creates Technology That Can Zoom In, Enhance Pixelated Images More

Google Brain Creates Technology That Can Zoom In, Enhance Pixelated Images

Comments Filter:
  • Google can put together images based on smaller images that look like faces.

  • No CSI (Score:5, Informative)

    by Thelasko ( 1196535 ) on Tuesday February 07, 2017 @05:51PM (#53822373) Journal
    I don't care how fancy the algorithm is, the original data was lost. This is still just a guess about the original content. It's just a better guess than was possible before.

    I just hope law enforcement doesn't think they can use this to solve any crimes.
    • Is this Blade Runner-esque? Decker summoned some wicked camera technology. Don't bother me with those pesky limits to the physical laws.

    • Re:No CSI (Score:5, Insightful)

      by rmdingler ( 1955220 ) on Tuesday February 07, 2017 @05:58PM (#53822419) Journal
      Law enforcement [innocenceproject.org] would never rely upon unproven methods to improve conviction rates.

      • Absolutely not. If that was allowed to happen then before you know it they'd be acting as an army (except with different coloured uniforms) and we all know that'll never happen [samuel-warde.com] because FREEDOM and NUMBER ONE!

      • Re: (Score:3)

        by wbr1 ( 2538558 )
        This needs so many more points. As someone who has been to prison - guilty as charged and pled to it, there are many who were not. There are plenty who were guilty in prison, but one that is not is too many.

        One particular case stands out to me. Navy soldier, on leave, drunk. Seen in an altercation in a bar with someone. that someone was known for instigating fights. Person winds up stabbed to death later.

        The suspect is arrested, and drunk, and with a huge lack of sleep, under duress from trained psy

        • So, the Navy goes to Great Lakes Naval Training Center for incarceration... were you also Navy or did you get your mail at the USDB?

          • Re: (Score:2)

            by wbr1 ( 2538558 )
            Not all crimes are prosecuted by the military. Sometimes the local or state authorities do, often with a discharge or additional military justice. I was never in the militar, but knew several in state prison that were at the time of their arrest.

      • Law enforcement [innocenceproject.org] would never rely upon unproven methods to improve conviction rates.

        That are good points. Law enforcement are supposed to use each of those evidence as leads. However, they tend to fixate on the what fits most to the case instead of thoroughly check out those leads. Once they have a suspect, they will try to fit as much evidence to the suspect as they can. They also eliminate the parts that don't fit (or not admit them to the court). As a result, it causes an innocent to be convicted.

        I don't know why they want to close the case ASAP. Possibly, it is like a trophy for how ma

    • The interesting thing would be if a system like this could actually recognize details from previously-seen better-resolution pictures of the same thing to actually fill in details in a justified way.

    • Why couldn't they solve crimes with this? It might not be fool proof but I see no reason why it couldn't narrow the search.

      • Sure, it could narrow a search. However, this shouldn't be used as evidence in a trial, or even to obtain a warrant.

      • Re: (Score:2)

        by aiht ( 1017790 )
        Yeah, no. If you want to zoom in on a tiny blur that you know is a face, this tech will give you just that: A face. A different face that it found in a different (higher-res) photo. That can't possibly ever do anything other than mislead.

      • Re: (Score:2)

        by skids ( 119237 )

        Because it is not image enhancement, it's image embellishment. It would quite often "narrow" the search onto the wrong track whenever it encountered an atypical situation, and in court it could be claimed that its results were only a convenient pretense, not a probable cause for suspicion (an authority could shop multiple such competing services until one hit on someone they wanted to harass.) The shit would really hit the fan when the algorithm was inevitably shown to have some significant accidental bia

    • Re: (Score:3)

      by PPH ( 736903 )

      I just hope law enforcement doesn't think they can sell this to a jury.

      FTFY.

      CSI has already proved problematic in that jurors have developed unreasonable expectations of what is possible [fjcdn.com].

    • Comment removed based on user account deletion

      • Re: (Score:2)

        by AmiMoJo ( 196126 )

        We all root for the protagonists on CSI, Criminal Minds, Law & Order

        I don't. Especially those bastards on CSI. "Why don't you take a DNA test, rule yourself out of the investigation?" Fuck off.

    • Re: (Score:2)

      by msauve ( 701917 )
      But, Google.

      Who are you to say that something from nothing isn't possible? :)

    • It doesn't matter if law enforcement thinks it can solve crimes. What matters is if lawyers think juries will believe the results. Juries think that CSI is real and forensics can all be accomplished in less than an hour.

    • Re: (Score:1)

      by aod7br ( 573614 )
      I could be used for law enforcement, not to confirm a candidate, but to discard suspects that don't match.

    • The algorithm can paint anything it wants into those squares. Want a video of Donald Trump getting peed on? Just film a low res video and have the algorithm paint his face into the place of the actors.

  • Stupid (Score:1)

    by Anonymous Coward

    This is just pure, 100% guesswork done with a computer. You cannot 'enhance' information that simply does not exist. The 10% of "fooled" people just mean those people were not familiar enough with what that celebrity actually looks like to tell the difference.

    CSI "enhance it!" remains fiction, sorry.

    • When I have my glasses off, I can recognize things my eyes can't resolve. Human stereo vision in relative depth perception depends on measurements that are more precise than the layout of cone cells would suggest. I can reliably get information that I can't see. Why can't a computer?

      • There are several things that help humans.One is that you can take several perceptions over the course of a fraction of a second and unconsciously merge them in a manner that improves resolution. Another is that humans can do pattern recognition of things buried in noise that a computer can't do unless specifically programmed to do that specific thing. (Think of resolving a head of hair in in a dim corner in a dim room.) Another is that the brain just makes up stuff (read about the blind spot.) You may be p

        • Sure. In many ways, I'm a far better computer than has ever been built in silicon. That doesn't mean that what I do can't be partially replicated in a powerful computer. I'm using myself as a proof of concept, that it is possible to beat the resolution, not claiming that any given system is or is not able to do the same.

  • Ideal test case (Score:4, Interesting)

    by fibonacci8 ( 260615 ) on Tuesday February 07, 2017 @05:57PM (#53822415)
    Feed it minecraft screenshots and japanese porn, and see what the result is.

    • Feed it minecraft screenshots and japanese porn, and see what the result is.

      I was thinking Super Mario Brothers sprites myself. Those were 8x8 sprites, if I remember correctly.

      As the article points out, this is less, "Zoom, Enhance" and more "Best Guess". Unfortunately, years of bad computer science on TV is just going to confuse people into believing this algorithm can do far more than it actually does. I wish Google would open it up so we could test it with our own images and show others how untrustworth

    • Well, if you are into that kind of stuff (neural network generated kind-of porn but not really), I suggest you check this out: https://open_nsfw.gitlab.io/ [gitlab.io] (Warning: somehow NSFW, or just weird, I don't know, but it will give you nightmares).

  • It's about time! (Score:5, Funny)

    by npslider ( 4555045 ) on Tuesday February 07, 2017 @05:59PM (#53822429)

    So in other words... from a small picture of the earth viewed from orbit, Google can now show me my house AND the address on the UPS package sitting at my doorstep?

    Amazing!

    • Comment removed based on user account deletion

      • There was once a company trying to sell infinite compression technology.

        Information theorists chuckled knowingly, and of course the 'inventors' failed miserably to decompress their compressed data... but at some time somebody actually thought it was possible and tried to bring it to market.

        • Black Hole Inc.?

        • That's easy. Here's a compressed version of the Library of Congerss: 1. I'm still working on the decompression algorithm. That will take a little more time.

          Back when Bruce Schneier facts were a meme, one claimed that, when Bruce Schneier wanted to write a book, he generated a random string of the correct size and decrypted it.

      • "I can see my house from here!" ;)

        (Simulation of Kronos, the Klingon capital city. Star Trek Enterprise Season 1 Episode 05 - Unexpected)

      • A friend of mine was involved in a vendor demo for compression to reduce the number of CD-ROMs needed to contain software. He insisted that the information was already compressed, and everybody was ignoring him until the vendor's software increased the number of CD-ROMs.

    • So in other words... from a small picture of the earth viewed from orbit....

      Only if Google already has a high resolution picture of your house with the UPS package already sitting on your doorstep.

      The algorithm does NOT enhance pixelated images in any meaningful way. It is only able to match a pixelated image with an already existing high resolution image of the same thing, and only by scaling the high resolution image down to a pixelated form suitable for comparison with the existing pixelated image.

      The only thing that makes this at all interesting is that pattern matching algori

  • What about zoom out ? (Score:4, Funny)

    by alexhs ( 877055 ) on Tuesday February 07, 2017 @06:01PM (#53822439) Homepage Journal

    If it doesn't do uncrop [dailymotion.com], it's lame.

  • Sorry, but the VT state troopers already have this technology: https://www.youtube.com/watch?... [youtube.com]

  • "...it was quite successful against humans..." (Score:4, Funny)

    by turkeydance ( 1266624 ) on Tuesday February 07, 2017 @06:06PM (#53822483)
    now we know the perception and purpose,

  • Biased selection? (Score:1)

    by Anonymous Coward

    So Google fed its algorithm millions of high-resolution images. It then feeds a pixellated image of a celebrity, and AMAZING! The algorithm can create an image of that celebrity so detailed that it fools people.
    Of course, chances are a good portion of the source material is pictures of that celebrity. This would be far more impressive if it could construct a police artist profile image from a pixellated source. One that could be reliably matched to an individual that's not even in the massive collection of

    • When you think about what they'll probably do with it, this makes perfect sense. If you've ever used Google Images' "Visually Similar" feature, it doesn't work all that well - especially when you're trying to find a higher resolution image.

      If they can do a pseudo-reconstruction, they're partway there.

  • no need to photoshop (Score:3)

    by zlives ( 2009072 ) on Tuesday February 07, 2017 @06:12PM (#53822539)

    this is perfect for your "real" profile pic on dating sites, just upload a google enhanced image of your self created from your 8x8 pixel image. yes this celebrity is really me.

  • They've been watching too much TV (Score:5, Insightful)

    by taustin ( 171655 ) on Tuesday February 07, 2017 @06:14PM (#53822551) Homepage Journal

    TV Detective: "We have this security video showing the murder."

    TV Lab Rat: "It's too grainy to tell how is it?"

    TV Detective: "Can't you enhance it?"

    TV Lab Rat: "Sure. Who do you want it to look like?"

  • This is nothing new. They have been doing this in movies for ages.

  • Really impressive progress (Score:4, Informative)

    by PhrostyMcByte ( 589271 ) <phrosty@gmail.com> on Tuesday February 07, 2017 @06:23PM (#53822615) Homepage

    A while ago, someone made the nnedi [doom9.org] upsampler that uses neural networks to upsample. It's still one of the best image upsamplers available.

    Google's approach is quite a bit different. Where nnedi worked to better extract detail out of what was already in the image, Google seems to literally fill in detail that was probably in the source but maybe not. Much, I guess, like how our own memories work. It's an interesting approach and the results look quite fantastic. My only question is how well it will work on a random sampling.

  • Move over, Mount Rushmore, Google now has an algorithm that can wallpaper the Ceres asteroid with the face of every American who has ever been photographed—all the way back to a pinhole camera exposing an onion skin soaked in lemon juice and potato starch.

  • The only reason this works as well as it does (poorly - 10% to 28% according to the post) is that humans are hard-wired to find faces among images. We'll accept anything with two small ovals on the same line with another shape roughly in the middle, below. Nose optional. Try the equivalent with random images or music, and I'll bet the success rate would drop to nil.
  • Fractal Image Format was compression that used self-similar parts of the same image to decopress to higher resolution https://en.wikipedia.org/wiki/... [wikipedia.org]
  • In the 1980's people were trying to this with fractals. I remember reading about something very similar and wish I could have afforded the book. I wonder what google is doing. Probably brute force.

    I was really interested n fractals and fractal compression for a time. And while you could get some insanely high compression ratios, the technique was lossy and decompression took 50+ hours (and 2 hours to compress). A potential use of the technology was picking out interesting artifacts from low rez space oho

  • Bad summary (Score:5, Informative)

    by swillden ( 191260 ) <shawn-ds@willden.org> on Tuesday February 07, 2017 @06:51PM (#53822717) Journal

    The summary's explanation of what this does isn't correct. It says:

    Google's new software effectively extracts details from a few source pixels to enhance pixelated images.

    It doesn't extract details from a few source pixels. It invents details to add to those source pixels, based on the knowledge that the pixelated image is of a face, and of what faces look like. It produces something that plausibly fits the input data. How close this is to the original image, pre-pixelation, depends on what images were in its training set.

    This is an interesting piece of work, but it doesn't mean that you can recover data that has been discarded.

    • ^^ this

  • On a single image the additional detail is guess work, the information isn't there. But think multiple slightly different but related images of the same subject, like any video filmed in the past 100+ years. Now you can cross-check and improve guesses with time correlation, and really find things that are there in terms of information, but invisible to the eye because each bit is spread over multiple frames.

  • They do that in movies all the time (Score:3)

    by OneHundredAndTen ( 1523865 ) on Tuesday February 07, 2017 @07:26PM (#53822867)
    Does this imply that movies have been lying to us all along? :-)
  • When I post photos that have personal information, license plates on cars, house numbers, names etc, I will do a two fold thing. Use the lens blur to blur them out, then go back over them with a masking color. Pretty simple in photoshop.

  • Come on Google (Score:4, Insightful)

    by CODiNE ( 27417 ) on Tuesday February 07, 2017 @08:58PM (#53823281) Homepage

    At least make the world a better place...
    by making this a MAME scaler.

  • For some reason, the first thing that comes to mind is a 2D pixelated Luigi now transformed into a six packed ultra good looking guy with the most macho mustache ever.

    Maybe I need more coffee.

  • > 8x8 pixel images which it then turned into some pretty clear photos where you can actually tell facial features apart. This is bullshit. It is mathematically impossible to compensate such loss of information. They might use other images to reconstruct original form a pattern, but this, obviously, has nothing to do with reality if the other images used as sources of details have not been made at the same moment.

  • If they have, say, a minute of pixelated video presumably they could estimate the orientation and position of key features of the face and then make progressively improving estimates of a higher resolution image.

    I know that was not the focus of this research (to match a pixelated image to one of a number of high resolution alternatives). But much of the crappy blurry images I see are in the form of video and it seems to me that there would be multiple independent images of a face that could be assembled in

    • That's a very interesting idea/application of the tech. Nice work. All those crappy convenience store robbery videos might actually be of help after that.
  • If there are multiple low level images such as would occur in a surveillance camera. It should be possible to combine them a get a much better guess at a true image.

Slashdot Top Deals

"Confound these ancestors.... They've stolen our best ideas!" - Ben Jonson

Close