back to article Google plonks reCAPTCHA on Street View, makes users ID your house

Google is getting the public to identify house numbers and signs from Street View photos as part of its reCAPTCHA anti-spam technology - and feeding the data into its online mapping service. Reg reader Jim Allen was first to notice that photos began appearing in the reCAPTCHA tests over the last week or so. These images are …

COMMENTS

This topic is closed for new posts.
  1. This post has been deleted by its author

  2. Lee T
    Big Brother

    Not Actually A Change

    you've been able to tell the known and unknown text apart for a while - the known text (currently) always has a section inverted, so theres a black splotch with the text in it turned white. previously the known text would be repeated behind itself but twisted a few degrees.

    1. Old Handle

      Re: Not Actually A Change

      Yeah, that has been true to various degrees for quite some time. Even before this last change it had become super obvious. I used to enter garbage for the unknown part on principle, but eventually I realized you can just completely ignore the non-scrambled word.

  3. Bakunin
    Terminator

    Changing the control part

    But assuming the high volume of CAPTCHAs that Google are providing to other services, they'll quickly reach the point where they've identified a significant number of unknown images to be able to use them as known control images.

    However, what I don't like is the number of CAPTCHAs I've encountered where *I* can't work out what they are let alone a machine. Perhaps I'm just not human enough?

    1. Jamie Kitson

      Re: Changing the control part

      They wouldn't use them as control CAPTCHAs, that's not the point. What they will (presumably) do is to have multiple users validate the same text.

    2. Anonymous Coward
      Anonymous Coward

      Re: Changing the control part

      I noticed this too, the increasing number of unreadable captchas. Makes me feel sorry for the less computer savvy (eg. the elderly) on the internet who might give up on whatever form it is (and often blame themselves that they couldn't work it :( ).

      Of course there is a speaker icon but those new to the internet wouldn't think to use it.

      1. I ain't Spartacus Gold badge

        Re: Changing the control part

        Not just the less computer savvy either. I really struggle to read captchas, and almost never get them right on the first try - due to dodgy eyesight. I've tried a few of the listening captchas, and couldn't make head-nor-tail of them because of the incredibly loud background noise, and quiet computer generated voice. Not all office PCs have speakers anyway - this one just has incredibly crap ones.

        I guess this is one of those cases where the 'mild' inconvenience to a few people may be better than the worse inconvenience of spam to everyone. Although it doesn't stop me swearing at them sometimes.

        Maybe a cartoon would be better, as spammers keep getting better at text recognition.

        1. Anonymous Coward
          Anonymous Coward

          Re: Changing the control part

          Good points Spartacus. I have a good friend who is partially sighted, wouldn't have a chance, can barely see the speaker icon anyway.

          Did someone not once take out a patent on a capcha, but based on animals?

          ie. A cat, a mouse or a dog?

          The reasoning being that automated systems find it difficult to differentiate between animal pictures.

  4. LJRich
    FAIL

    Not Exclusive, Sorry

    TechCrunch got there first, last week.

    http://techcrunch.com/2012/03/29/google-now-using-recaptcha-to-decode-street-view-addresses/

  5. Anonymous Coward
    Anonymous Coward

    What's to stop someone from deliberately labelling a house number wrongly? if Google doesn't even know it, they can't claim it's incorrect!

    1. DZ-Jay

      Even if every user enters the wrong text for the unknown image, I would imagine that they have a very low chance of entering the *same* wrong text. Therefore, the image never gets a high enough confidence of correctness. The end result is that there is no change in state for the image, it remains unknown.

      That may not help solve the unknown images, but it doesn't necessarily pollute the database with bad data.

      -dZ.

      1. Audrey S. Thackeray

        Yes, it would take a large co-ordinated effort to pollute the data - and it would probably take something more sophisticated than 'enter the numbers backwards' or 'add one to each number'.

        1. Steven Roper

          A couple of years ago

          4chan ran a campaign to replace the "unknown" word in captchas with a well-known anti-African racial slur. As I recall, they pushed it for several months - but I never saw the slur nor any effect from it either in captchas or in Google Books as a result. Even with 4chan pushing it, I consider it highly unlikely that that their numbers made a significant difference against the huge numbers of people who at least attempted to get it right.

          And even if there were enough trolls to make a difference, I'm sure the Oompa-loompas would pick up on it sooner or later and simply erase the effect of anyone entering that word - after all, they would already have made sure that racial slurs and other un-PC words wouldn't make it into the captcha system anyway, for fear of lawsuits.

    2. Colin Miller

      Except that the same image will be served to several (possibly many) different people. If they (almost) all agree, this will be taken as the correct number, otherwise an Oompa-Loompa will have to manually tell the googleplex the correct number.

    3. dotdavid
      Facepalm

      They show the number to multiple people rather than taking one person's word for it.

      1. Anonymous Coward
        Anonymous Coward

        If the user who is deliberately using incorrect numbers, starts using proxies and other IP/browser combinations to hide their own house number, it will quickly backfire on Google.

        1. Anonymous Coward
          Anonymous Coward

          Chances?

          What are the chances of even a bot net getting 2 of the same images to try and confuse the system?

          Plus, why? Other than to troll? It's not the active part of the Captcha. So they take out door numbers, how you scam from that?

    4. Anonymous Coward
      Anonymous Coward

      Why do the incorrect labeling that way? I'm going to wait for the Google car to be in my neighborhood, and tape over the street signs with my own label. After a few thousand people do the captcha, I'll be living on 'Bite My Shiny Metal Ass Drive'. Who's with me?

  6. NozeDive

    I've been wondering what those blue squares were. I always thought they were from a book, like the other reCaptcha images.

    It should be noted that those house number images, if that's what they are, have been showing up in reCapptchas since 2011. I guess it's news worthy if Google just made the announcement.

    If you're not happy with identifying house numbers, you don't have to. Just correctly type the word used for identification (you'll be able to tell them apart after doing enough of them) and put gibberish of an appropriate length where the scanned image word (or house number) goes.

  7. Anonymous Coward
    Anonymous Coward

    Re: Incorrect labling

    Re: What's to stop someone from deliberately labelling a house number wrongly?

    Maybe they don't rely on one individual and rely on an average sample based on results from around the world.

  8. 2-Part Return

    Bit longer than a week

    I got one of those in November. Number "190" on the same blue plate style thing.

    stop spam. read books. update coordinates.

  9. Jamie Kitson

    So...

    Deliberately get that part wrong?

  10. JetSetJim
    Meh

    What about the deaf folk?

    How does the reCaptcha make audible the unknown number? Yes, once it's had a significant number of people say it's "221B" it can then read it, but if it hasn't yet reached that statistical threshold?

    I guess it could move on to the next one if the user presses the "listen to" button. Or just read out the control (in which case it is given away as to which word is the unknown in this case)

    1. D@v3
      FAIL

      Re: What about the deaf folk?

      "How does the reCaptcha make audible the unknown number?"

      How is reading something out supposed to help deaf people?

      1. Audrey S. Thackeray

        Re: What about the deaf folk?

        A mate used to have a near-deaf colleague who had a hearing dog (like the guide dog for the blind idea). Every time her phone rang the dog would bark its head off.

        Obviously she couldn't hear the dog any better than she could the dog (&bone) - and even if she noticed it she wasn't very good on the phone even with an induction loop for her hearing aid.

        Like a flea-carrying version of something I saw not on the nine o'clock news once.

    2. JetSetJim
      Facepalm

      Re: What about the deaf folk?

      Doh. Posted in a hurry

    3. Dragon Dave
      Megaphone

      Re: What about the deaf folk?

      The listenable captcha is completely independent of the text-based one; last time I checked it was a sequence of numbers read out over a noisy background of chatter.

    4. Anonymous Coward
      Anonymous Coward

      Re: What about the deaf folk?

      The Thumbs down trolls are at it again!

  11. Anonymous Coward
    Anonymous Coward

    Re: What about the deaf folk?

    How does the reCaptcha make audible the unknown number?

    Have you ever done any conditional programming scripting ?

    $v_deafnum = 2;

    $v_deaf_button_pressed=FALSE;

    function press_deaf_button {

    $v_deaf_button_pressed=TRUE;

    }

    if ($v_deaf_button_pressed) {

    expect $v_deafnum ... blah blah blah

    } else {

    make a note of what human OCR told me

    }

  12. Anonymous Coward
    Anonymous Coward

    by the way...

    yes, I left it to the commentards to debug ;-)

  13. Anonymous Coward
    Anonymous Coward

    How is reading something out supposed to help deaf people?

    Shouldn't that have read ...

    "How is reading something out supposed to help deafblind people?"

    Otherwise you'll just get the usual commentard answer of... "well, they can see it, can't they ?"

    1. D@v3

      Re: How is reading something out supposed to help deaf people?

      that's as may be, but it isn't going to /help/ them, is it.....

  14. Irongut

    The tech is mostly used as a challenge-response test to frustrate the automated sign-up to web mail accounts and similar services the hell out of me because I find them very hard to read and usually take at least 5 goes to get them right!

    Fixed that for ya.

  15. earl grey
    Unhappy

    they're a pain

    I had a site that was using some version of capcha and i didn't realize i had to have scripting turned on for it to work....so i tried the audio....and i couldn't understand what they were saying on it, so i moved on down to the idiot's version where it said, "type this in"....and it finally worked... so i just use the idiot version all the time...less hassle and frustration

  16. TeeCee Gold badge
    Mushroom

    "..reCAPTCHA users don’t know which word is which."

    Hint: The "unknown word" is the one spelled "rscintunf" or similar.

    I had sort of wondered why these things popped up with a word and a rejected new car name from the nearest Strategy Boutique. I'll save myself the typing and just enter the obvious control word from now on.

    I do find that ReCaptcha does seem to generate the most unreadable ones of all the types out there. Certainly the only ones that give me a consistant "miss" rate. I can see why Google wanted 'em, a fiendishly clever algorithm that produces useless shit is right up their street (Living at number 43a actually).

  17. Mike Brown

    genuis

    this is really clever on googles part. they get the general public to fill in the blanks on streetview. how ever came up with this idea needs a haerty slap on the back.

  18. Paul Hovnanian Silver badge
    Facepalm

    Re: Incorrect labling

    Fine. Until you call for that ambulance when you're feeling short of breath.

  19. Dave Bell

    Other unreliabilities

    If you're an Opera user, don't bother trying to comment on a blog that uses ReCaptcha.

  20. Oldgroaner
    Thumb Down

    Evil, evil, evil.

    1. Anonymous Coward
      Anonymous Coward

      Curse those maps and their privacy invasion! Next thing you know, UPS, Fedex, and the Postal Service will know *EXACTLY WHERE YOU LIVE!*

      Shudder.

  21. Grease Monkey Silver badge

    They'll have fun on our street then. Most of the houses don't display their numbers. Those few that do display them in strange way. Number 14 has it's number on the garage and due to the position of the garage most people assume it belongs to the house next door. So number 16 gets plenty of visitors and deliveries for number 14. If Google use their software on this street it will make matters worse rather than better.

    1. Anonymous Coward
      Anonymous Coward

      The fact that it will potentially decrease accuracy for #14 and #16 in your neighborhood has no bearing on whether it will be useful over all. Package delivery is generally quite accurate - a few outlier neighborhoods with crazy-ass address plates are hardly a death blow to something designed to work with the other twenty million normal ones.

This topic is closed for new posts.

Other stories you might like