Gazopa raises the question: Is image search a feature or a product?
Hitachi research project unveiled. It's impressive, but is it too late?
Following the GazoPa took the stage with a similar product, but for photos.at TechCrunch50, the team from the Hitachi project
GazoPa is a pattern matching service. It matches color and shape. Or -- and this is cool -- just color, or just shape. Found a picture of a product you like, but want to see it, and only it, in other colors? Filter by shape. Or for similar products in the same shade, filter by color.
The presenter said the database handles 50 million images. But while it recognizes faces, it can't tell one person apart from another.
Cool feature: You can sketch a shape, and find images that match it. Or, more realistically, upload an image to the service and find similar results from around the Web.
The service will also search within videos, although it doesn't have quite the chops of VideoSurf.
The beta is due to launch today.
As a research demo, this was a cool project. But as a standalone product, it's a questionable business. Like.com has a similar enterprise. And as TC50 judge Bradley Horowitz from Google pointed out that this type of thing is more a feature than a product. Google's own Picasa Web just launched face recognition as a feature.
I agree with Horowitz. Eventually this feature will be baked in to search. But perhaps Hitachi can license its visual engine to the search companies that want to do it.