Google Image Recognition

[sgietz]

Our boss mentioned something about Google working on some sort of image recognition search. You could snap a pic with your cell phone, upload it to Google, and if G recognizes it, would tell you more about it. I tried looking it up, and I did find some articles on that, but they were about a year or two old. Does anyone have more recent news on this?

That has great potential. Imagine you're on the road, and you're getting hungry. You could stop and snap a picture of "Mom and Pop Country Restaurant", upload it, and it would give you ratings, reviews, a menu, etc. Of course my example is probably wishful thinking, but the potential is there.

[louieramos]

It would be good if they are working on that concept, but its probably not for search, because its easier to type than to upload photos and for google to process and recognize it. They will probably have a different purpose for it if they are working on that technology

[caribguy]

The panoramio pictures are now meshing: it lets you navigate from one to the next. Saw this for the first time today.

[briggidere]

It's called Google goggles. It's available for Android phones and I had a demo of it when our Google person came to see us a few months ago. We just took a photo of the Eiffel Tower on the screen of a pc and it picked it up and told us more about it. Quite cool. They are apparently trying to develop it for the iPhone as well.

Have a look at [google.com...]

[PhotonLight]

I tried it out a bit. It works ok for specific things, major landmarks, anything with a barcode, etc... but still pretty rough around the edges and lots of stuff it doesn't work with. It's a good start though.

[Robert Charlton]

There's some discussion of Google Goggles, as well as Google Similar Images in this thread....

Google Goggles - search by submitting a photo
http://www.webmasterworld.com/google/4039028.htm
[webmasterworld.com...]

Explanations of image recognition software (and image stitching software) that I've read suggest that an image can be broken down into polygons, and then matched with other images using "corners" of shapes defined by the software.

It works ok for specific things, major landmarks, anything with a barcode, etc...

Clearly, you need a starting point for matching. You'd otherwise have to crunch an impossible amount of data to find a match. The more data you have, also, the more refined and accurate the entire model becomes.