Follow Slashdot stories on Twitter


Forgot your password?
Communications Supercomputing IT

Cellphone App Developed that Could Allow For 'Pocket Supercomputers' 73

Jack Spine writes "A robotics researcher at Accenture has given a demonstration of a 'Pocket Supercomputer' — a phone behaving like a thin client. It can be used to send images and video of objects in real time to a server where they can be identified and linked to relevant information, which can then be sent back to the user. 'The camera on the phone is used to take a video of an object — such as a book ... By offloading the processing from a mobile device onto a server, there are few limits on the size and processing power available to be used for the storage and search of images.' To pinpoint the features necessary to identify an object, the image is run through an algorithm called Scale-Invariant Feature Transform, or SIFT, a technology developed by academic David Lowe. The software extracts feature points from a jpeg and makes a match against images in the database. If a match exists then the software on the server retrieves information and sends it back to the user's phone. A 'three-dimensional' image of an object can also be uploaded onto the phone, to look at the virtual object from different angles. The motion-tracking technology Accenture uses for this is a free library of algorithms called Open Computer Vision."
This discussion has been archived. No new comments can be posted.

Cellphone App Developed that Could Allow For 'Pocket Supercomputers'

Comments Filter:
  • All boobies examined with the low quality phone cam will resemble Lara Croft.
    Hell, it might make the same distinction if you take a picture of some melons.
    • Re: (Score:1, Funny)

      by Anonymous Coward
      Is that a supercomputer in your pocket or you just happy to see me?
  • Didn't someone already do this?
    • Actually, yes. This is not new. There are already many robots out there that use this exact same technology. The only difference is that this guy used a cell phone.

      This is no more "innovation" than a port of an existing software application to a new hardware architecture is "innovation."
  • by Lisandro ( 799651 ) on Thursday January 31, 2008 @09:49AM (#22245550)
    By offloading the processing from a mobile device onto a server, there are few limits on the size and processing power available to be used for the storage and search of images

    That's like saying my TV set at home can be called a miniature television studio.
    • by somersault ( 912633 ) on Thursday January 31, 2008 @09:56AM (#22245606) Homepage Journal
      I concur.. shouldn't this be "cellphone thin client developed"? Or is that not groundbreaking enough for the mods..
      • by Serapth ( 643581 ) on Thursday January 31, 2008 @10:13AM (#22245780)
        I agree completely, the title is highly misleading. The cyclical nature of our industry does make me laugh though, its like every few years the balance of power between the network and the device shifts and WOW!!! a whole new way of computing!!! Repeat and rinse. How people fail to recognize its the same thing over and over, boggles my mind. Oddly though, the cell phone has basically been a thin client all along, and its only recently with Palms Treos, Windows smartphone, Apple iPhone's, etc... that things started trended towards thick clients again.

        Lastly, atleast here in Canada, this idea is completely unrealistic anyways as the bottleneck is essentially the network not the device. A combination of high data charges, no flat rate billing plans and slow networks just doesn't mix well.
        • Nah it's the same in the UK as well, and most likely all other countries. Data prices are completely insane.
          • It's getting better... you can get a gb/month or so (subject to the usual restrictions, like no voip, msn, etc.) whereas a couple of years ago you were lucky if you got 4mb/month.

            I still remember the £300 bill I got from connecting my laptop to the phone one day and leaving windows update still enabled.. aargh. That was when the excess was something like £5/mb.

            Mobile data will become useful to me when it's at about the rate of DSL.. or even in the same ballpark would be nice.
        • Canada is slowly getting better though. It wasn't that long ago that a $100 data plan with Rogers earned you only 100mb of wireless downloads. Now with the newly unveiled Flexrate data plans (mainly intended for their hsdpa laptop data card users) you can get 5gb of downloads for $100. That's definitely not garnished with an 'unlimited' label, but they're definitely getting closer. (more info: [] )
      • IMHO, "cellphone thin client developed" isn't really ground-breaking. Opera Mini [] is an example of prior art. Now, granted, it ain't supercomputer-like, but still, the concept is similar...
        • Exactly. You also get Remote Desktop/Terminal services on WM smartphones. If you're gonna say stuff like that then you could say that anything connected to a network is a potential 'supercomputer'
    • I was looking for the right analogy, and that one's good enough.

      I just tried to tag it nothingtoseehere and it said: "If you disagree with hingtoseehere, please use !hingtoseehere instead. You can edit this text field in place and click 'Tag' right now.". So let's hope we never get any articles about The Sherrif of !tingham.
      • by Barny ( 103770 )

        I was looking for the right analogy, and that one's good enough.

        Nah, it needed to use pipes or cars :)

        The biggest problem I see for these things, bandwidth. I will have no problems at all as long as my phone doesn't leave my 802.11g AP range, but outside that...

        Heres a good story, I live in Australia, I have a nice shiny 3g nokia phone with all the trimmings, I have all data service providers except for my home AP removed from the settings. A friend gets a problem with his sim card, I test it in my phone (y

    • I can see the 1969 headlines now... "Entirity of United States of America emigrates to the moon, comes back" ;)
    • Seriously. If the ability to talk to a "supercomputer" makes a cell phone a supercomputer, then I'm a supermodel.

      Marketing note: Apply the laugh test. Tell someone who actually works with supercomputers that your cell phone app makes a "pocket supercomputer". If milk comes out their nose, don't run the story.
    • Umm... bad analogy. TVs only display the pictures. A TV studio sends them. Your TV doesn't send video, does it?
  • ... the Borg system?
  • Uses for the blind (Score:3, Interesting)

    by IndustrialComplex ( 975015 ) on Thursday January 31, 2008 @09:59AM (#22245634)
    Wasn't there a segment on NPR about a cellphone with an integrated 'reader' that did OCR on pictures taken with your cell phone? I could definately see OCR as something that you would want to offload to a server, potentially one designed just for that.

    • by 3waygeek ( 58990 )
      My HTC Tytn II phone has a built-in app called WorldCard Mobile that does something similar. You take a picture of a business card, and the software does OCR on the image & creates a Pocket Outlook contact record with the OCR text. It's not the best OCR I've seen, but it's not the worst, either.
    • the server software is smart enough to recognise the cover of the book -- it's not yet able to read text -- and can then, for example, return the price and history of the book, and details of where it can be bought.
      The whole point of their thin client is to push your consumerism to the next level.

      Hooray for innovation!!1
    • Already available at []
  • The age old question (Score:4, Informative)

    by Pojut ( 1027544 ) on Thursday January 31, 2008 @10:00AM (#22245638) Homepage
    We really need to stop throwing the supercomputer term around. How do you really define supercomputer? Is it based on number of calcs per second it can do? Size? Hell, my PSP has more power in it than room-filling monstrosities from the 50's...
    • The Dept. of Commerce [] defines it as a machine with a composite theoretical performance equal to or exceeding 1,500 million theoretical operations per second.
      I, for one, find this definition highly stupid and refuse to call a machine capable of less than two orders of magnitude more FLOPS than "enthusiast"-grade hardware of the respective timeframe a supercomputer.
      Even though your PSP may outperform them, "the room-filling monstrosities from the 50s" remain supercomputers to me while a '08 supercomputer wo
    • I agree. We need to stop misusing these terms.

      Plus, I'm concerned this software could brick my cell phone, requiring me to reboot.
    • 40 to 400 Teraflops at current numbers.
    • by PDX ( 412820 )
      If a computer is linked to others and performs better than original specs indicate that it should, then maybe it earns the elite term of super. If I call a latte grande but its only half full it doesn't get called supersized. If clever compression algorithms condense and extract data from a device, ie.., USB Flash, Hard drive, Rewrite CDs, or math coprocessors in PCI slots then you have the possibility of doing more with less. If you add networking you still have to deal with buffers, transfer rates, and er
    • Accenture?? Isn't that the elections-rigging specialist that screwed up on the porno premier of Italy, Berlisconi, in his reelection campaign??

      Yeah, if I recall, he hired them and he only "lost" by 25,000 votes while the exit polls clearly showed he really lost by well over 1 million. Guess Accenture screwed up on the details......

  • Or you could open the book, and turn its pages yourself
  • ...or are you just happy to see me?

  • Funny, I recently read a similar idea in a SF short story in Asimov's*. The character had built a supercomputer out of dumpster-dived wifi-enabled smartphones. Mesh networking, voice recognition, it's all included.

    Anybody wanna try implementing it? :)

    *Hormiga Canyon by Rudy Rucker and (?). Can't remember authors exactly, and Asimov's magazine website doesn't provide a bilbiography. In fact, their whole website is pretty shitty by today's standards. For an SF magazine, the irony is terrible.
  • While this sounds cool, it interests me more because of the fact that the Open Computer Vision algorithms [] are open sourced and in this case by Intel's research groups. While I might prefer if another microprocessor company was more dominant, there are some areas where Intel's interests diverge from the unholy Wintel alliance, and in these areas they do some really good stuff.
  • Barcode (Score:3, Informative)

    by Taulin ( 569009 ) on Thursday January 31, 2008 @10:39AM (#22246108) Homepage Journal
    Something sort of similar already seen in Japan. Look in any Japanese magazine, and in almost every add you will find a 2D square barcode. Point your cell phone camera at it, and it will look up the information. Basically, it is just a URL, but it is a standard thing over there. Really nice.
    • by Isao ( 153092 )
      I agree that QRCode [] is nifty, and hope that it and/or similar systems take off around the world, but this is a little different.

      All the QRCode processing is done on-phone. This idea has a tightly-coupled client-server relationship, and is a step in the direction of distributed mobile code and data (overdue and welcome, in my book).

      My biggest interest is how the trust model will work - if you subscribe to an image-processing service like this, do they own the picture, search metadata or profiling info? (

    • []

      "A QR Code is a matrix code (or two-dimensional bar code) created by Japanese corporation Denso-Wave in 1994. The "QR" is derived from "Quick Response", as the creator intended the code to allow its contents to be decoded at high speed. QR Codes are common in Japan where they are currently the most popular type of two dimensional code.

      Although initially used for tracking parts in vehicle manufacturing, QR Codes are now used in a much broader context spanning both commerci
  • I usually keep my phone in my pocket. After even a short time, I bet they will have tons of data to analyze about the inside of my pocket. This is important research.
  • by serviscope_minor ( 664417 ) on Thursday January 31, 2008 @10:43AM (#22246154) Journal
    It's just a thin client on a mobile phone, and not an especially interesting one. They use SIFT, so they don't stand a chance in hell of doing anything useful on the phone. There are other, modern, much more lightweight ways of identifying features. It would be more interesting performing as much processing on the phone as possible (for instance to reduce bandwidth and/or latency).

    Other than that, it's a neat hack.
    • It would be more interesting performing as much processing on the phone as possible (for instance to reduce bandwidth and/or latency).

      However, a cell phone has hardly the processing power to be able to perform image classification, let alone against a database of trained patterns. It might be able to extract one or two features from an image, such as RGB concentrations and hue/luminescence, but today's cell phones would have great difficulty extracting text from an image -- something that Accenture's "hack" can't yet do.

      With the diminishing costs to maintain back-end servers, or the popularization of volunteering CPU time toward distr

    • Could you give a quick list of those? I would love to read up on them.

      • Re: (Score:3, Informative)

        Sure! From the wiki [] (not a great article, but a good place to start): the best bet is the FAST corner detector. Also on the wikipedia page, there's Trajkovic and Headly with their paper "called fast corner detection", which is different from FAST corner detection". It's very similar, so it might be as fast if it used the same tricks. There's also the venerable SUSAN detector which is somewhat older and slower, but still one of the quicker ones.

        The wiki page is a
        • by wurp ( 51446 )
          Is corner detection in the same league as far as doing object recognition in a real environment as SIFT?
          • SIFT is part of an object recognition system. Sift finds "interesting" points, and computes useful features from the pixels around those interesting points. An object recognition system would then do something useful with SIFT points, such as matching in to a database.

            SIFT relies on finding interesting points to start with, and the field of finding interesting points is "corner detection". The standard SIFT implementation uses a difference-of-gaussian detector which is pretty slow.

            IOW corner detection is re
            • by wurp ( 51446 )
              Excellent point. I was looking at the corner detection as an alternative to SIFT, and I didn't see how that could be the case. I can see how it could be a great addition, though.

              Thanks again!
  • A 'three-dimensional' image of an object can also be uploaded onto the phone, to look at the virtual object from different angles.
    Have we lost the ability to turn or walk around objects to see them from different angles in real life?
  • So if I understand the story someday I could use my phone and home supercomputer to identify a book. "Hey what's that?". "Hang on a minite". Click, beep, beep, beep, ring, ring, wirrrrr.... "It's a book! And I can buy it on Amazon, what a suprise!"
    • Or maybe you could walk around looking through smart glasses, with many things you see enhanced by unobtrusive tags. When you shop, it can tell you which items are made in sweatshops, or are cheaper somewhere else you plan to go later, or have a recall notice. When you drive, it can flag cars that have been identified as risky drivers. When you deal with people, you can see whether other people like you have identified them as trustworthy.

      On a phone it's a little more limited, but at least you could get
  • Accenture has people who do research?
    My experience with Accenture is that they rob financial institutions by claiming to sell computing and management expertise.
    Instead, they bill junior people learning Word at $1000/day, because they are Accenture (with a capital A) and the principal who made the deal went to school with some high exec at the firm being fleeced.

    If there is any actual expertise at Accenture, that is indeed news.
  • *watches the evolution of Earth's first true AI's optical system*
  • I had no idea [] had robotics researchers. When I was working there, developing anything new was to be avoided if at all possible.
  • the supercomputer. It has millions of (semi-)intelligent processors craniofacially attached to its nodes. Though parallelism is rampant, effective clustering and resource allocation are rare to be found, that with all those processors just idly blabbering away at each other most of the time...

    With apologies to Sun, they were there first.

Things equal to nothing else are equal to each other.