Slashdot Mirror
Google Launches Search By Image
kai_hiwatari writes "At the Inside Search event being held at San Francisco, Google has announced a new addition to its search features — Search by Image. The Search by Image feature is something like Google's image search application for mobile devices — Google Goggles — but for the desktop."
copying http://www.tineye.com/ seems a bit seedy.
Have gnu, will travel.
That video was poor. It was full of animated graphics and few actual results. The actual results didn't even find many matches.
Also, how is Google's version better than TinEye's? Nothing they've said about Google's search goes further than TinEye's. I've used TinEye and it's worked well enough; much better than what was shown in that marketing video.
Google, I'm not impressed. You should do better.
They should call this technology point-n-grunt. Now I don't even have to worry about articulating my wants into words.
In all honestly thats pretty cool though, I've been expecting this for a while. Congrats on the cool new technology Google. *holds up picture of a a party hat and confetti*
If what I just said sounded like a troll, it was probably just a failed attempt at humor.
Huh, I wonder if TinEye will disappear under the might of Google? It's been pretty useful, especially the browser plug-ins!
It doesn't mean much now, it's built for the future.
This isn't search based on facial recognition, but on product recognition, or so it seems from what I've read so far. It will be interesting when you can point Google at a picture of someone and have it search for other pictures of the same person. Then check to see if you get different results with safe-search off. (I'm surprised that there isn't a dangerous search mode that only shows results that would be blocked by safe search.)
the most common use of this will be to find more pictures of that one girl you can't remember the name of from that pic you found online..... How long before someone uses this to identify, track, and hunt down a photobomber?
Great article. Not only does it get "TinEye" wrong (Tiny Eye? Really?) but it also fails to link to Google.
It's supposed to be part of images.google.com, although it's not working for me currently (the camera icon doesn't show up in the search box). There's help on how to use this feature here.
Do they keep the image and add it to their collection or do they toss it away?
Google is now doing "me too" stuff like MS. They are now officially no longer cool.
And suddenly the internets cried out in pain as the simultaneous upload of a billion boobie pics strained its' tubes to the core.
The Google demo video only showed places, art and animals, not people. As far as I know Google has the tech for that already, but not enabled due to privacy concerns. There is however PicTriev for face search, which however seems to be relatively low quality.
Yeah, Rob Malda's. You both are in the running for tiniest penis in the world. The results are too close to call even under our most powerful microscopes.
I think it will find many needles. Why do you need just one?
Needles? It will have trouble finding the haystack!
Google had the same functionality as TinEye in Google Labs for at least as long as TinEye has been around. They (weakly) integrated it into their main site in 2009 as a "Find Similar Images" feature. Google goggles (and this) does a lot more than TinEye, because it can find different images with similar content, while TinEye only finds the same image with minor cropping and filtering applied. And academia has been publishing papers on images search for years before either company made anything.
The difference is that TinEye found a niche business model for the (relatively) simple image search that it had, and developed it into a very useful tool for the limited capability it had. Google on the other hand, decided what they had wasn't good enough for their market, and kept working on it in the background until it was good enough.
Neither is a rip off of the other. They are just different approaches to different problems, both of which borrowed from prior research as well as adding their own improvements.
Sometimes you have a picture that you want to know more about, and so far the only option was to take a picture of your screen with a smart phone. A desktop based image search should be much easier to use.
:)
If being able to search using photos on your phone is already useful to a lot of people, being able to search from your home after you've stored your pictures on your computer just makes sense. I'm just eager to try it for myself since it hasn't seemed to have actually launched yet.
If you're into manga and that sort'a thing, you should know about IQDB, which finds an image based on a thumbnail on several image boards. So from a low-quality thumbnail you can usually get a desktop-sized image and a bunch of tags that tell you what it is. (And it's been around since the early naughties, although not at that url.)
Are they similar to Beer Goggles?
TinEye searches much more than exact images.
I just took a screenshot from Google Street View in The Museum of Modern Art. From the screenshot I cropped out a painting (and didn't even change the perspective) and searched at TinEye which resulted in this search. Colour me impressed. Once again, my image is just a screenshot from a photo taken non-orthogonally at a painting.
TinEye is also extremely useful to help understand a photoshop meme :-)
- Peter Brodersen; professional nerd
The site must be translated by Google Translate to other languages. I tried it with a few different language settings and the language seem to be total gibberish if you use anything but English.
The search site automatically show you pages that is auto-"translated" according to the primary language setting in your browser. It would be much better if it showed you pages written by a real human in one of the language present in your language list in your browser settings. As most people (except native English speakers) can read at least a couple of the largest languages, it shoudln't be that hard. [At least one of Mandarin, Portuguese, Spanish, Arabic, Russian, English, French, German or Dutch (or Scandinavian) is understood by most internet users, the languages is listed roughly in the order of universality (most of the people who understand Portuguese or Spanish live in South America or South Asia, there are much more people who can read French, German or Dutch living outside Europe, then in Europe, Arabic is the universal language of the Muslim world, Russian was the universal language of the East block (but learning English is more popular with younger people), most people who can read English, can read one of the other languages in the list better; there are many people who understand spoken Hindi-Urdu, Punjabi or Italian (more then understand English) but most of them can neither read, nor live in areas with affordable Internet access)]
Worse, you have to change your primary language setting in your browser to get readable pages, there is no easy link to follow to get pages in English. As the English language pages can be read by any 9 year old (and kids younger than that shouldn't be allowed to use this service without an adult supervisor) that have studied English for a year in school, while their native language counterparts is rather confusing (and set a really bad example for proper language use), this is really unnecessary.
I can see this being really useful for finding creatives commons licensed images. If you find an image or diagram you'd really like to include in a document, but it's not openly licensed, you could look for other images like it, but use Google Image Search's recently-added license filter to find CC-licensed materials.
Is this being done by a hash of some kind, or by some form of image comparison? The first method would be efficient but will only produce exact copies. I'm not aware of any variation on the latter method that isn't incredibly system intensive with large numbers of images...
and it crashed with the message "Number of matches exceeds allocated memory."
Dear Slashdot: next time you want to mess with the site, add a rich-text editor for comments.
At last, we can get an accurate goatse census.
I uploaded a photo of mine--not available on the net--and Google showed me six photos that are visually similar with differing subjects. I like it.
My hope for the future is that a photo I take of some unusual connector will return info on what it fits. Similarly I would like to identify other objects through their photos when they lack other info such as model numbers.
See a person you'd like to get to know better? Snap a pic, run through the service, and find their details.
Possibly more fun and profitable uses of matching random faces to names would be jury tampering, background checks for employees, finding Ex's of whomever you're dating, blackmailing married men leaving a strip club or adult book store, finding where the TSA agent who just groped you lives...
This is the news URL at digitizor.com:
http://digitizor.com/2011/06/15/google-search-by-image/
The same happens on today's other digitizor.com news at Slashdot (Mac OS X Lion Has a Browser-Only Mode)
-Ignacio Agulló
I've just done a test with (1) a small section of a public image of sushi, with all metadata removed and (2) a photo I just took of some bananas on a white background.
(1) correctly identified what it was, gave me websites using the full original image, full marks, 100%, v.g.
(2) Showed me chopping boards, boxes, a violin... anything where the image had roughly the same colours in roughly the same portion of the image.
So I conclude that it is both. The hash is extremely effective, as you might expect it to be with precomputed info, but the node search is rubbish.
You're only jealous cos the little penguins are talking to me.
Google will be sued for not matching all submitted images against copyrighted images to find IP thieves. Mark it down as a certainty.
Not an advert-ridden piece of shit hoping for all the clicks from Slashdot, which he didn't earn in any way, shape or form.
This is one of those Google features that I scratch my head over. It sounds great but is almost completely useless. I tried several images both from the web and from pictures of famous locations that I had taken on vacation. I tried product photos, and place photos. I tried the search page and the Chrome extension. I got one page crash (drag-n-drop an image to search by), and no results that could be considered anything but awful. The results were so far off the mark as to be silly. I posted a chair and got action figure toys. The one example of useful results and the reason I will keep it around is that is showed other sites that are using images that I have copyrights too but were captured from my site. I guess if I had the time I could go ask for some credit and get a little pagerank love.