GazoPa Lets You Sketch Your Search
We all love image search and I know barely a day passes when I don’t make use of it at least once.
Image search/processing technology has also been an area that has been going through a bit of evolution over the past few years.
Cooliris brought us a nifty 3D spacial image search interface. What The Font brought smart image processing to font identification and IDEE have given us some amazing visual search tools with Tineye and Piximilar.
Last week we saw a fresh player jump into the the image search game with the launch of the GazoPa public beta.
Much like Tineye, GazoPa is a service that analyses the content of each image — the shapes, colors, tones and lines that make it up — rather than just inferring it’s content based on it’s file name and the surrounding content as most traditional images searches do.
However, GazoPa also adds a few interesting new strings to it’s bow.
Firstly, GazoPa allows you to sketch a rough image directly into the search panel which it will then try to match.
Of course, that’s OK in theory, but does it actually work?
In my first quick tests, yes, it seemed to do a reasonable job. I started out with a fairly detailed and distinctive drawing — a quick sketch of my iphone — and the results were certainly good.
I then gave it a more advanced challenge for its second test case — a simple line figure of a dancer — and the results were more mixed.
The current algorithm seems to emphasize the general composition over specific shape and line configurations. As such, most of the results mirrored my (unimportant) off-centered layout, more than the more important positioning of the figure.
That said, the fifth image presented on the results page (above) is actually a very good match.
I tend to think my second test is a more likely real-world use case, but this *is* a beta, so we’ll have to wait and see how things develop as they accumulate more data.
GazoPa have also released a free iPhone app to compliment their standard web service. The application lets you use match images created via either the camera or finger doodles.
If this works *well*, it’s quite simply a ready-made, stone cold, killer app for them. Can you imagine being able to identify and price an antique lamp with a quick pic on your phone? Or identify a dog breed? Or the architect behind a building. The mind boggles at the possibilities.
Throw in a Firefox extension and standard URL and keyword based searches and Gazopa have certainly come ready to play. It’s still raw, but looks promising.