At Present, the systems works on paintings, carpets and coins; though they hope to extend the system to work on 3-dimensional objects soon.
Nishimura seems to be a fan of the idea:
Sounds like a pretty simple and practicable idea, patching together well-established technologies. Take a database of images of stolen artworks, and search it using other images and a pattern-matching application. You'll end up with some false positives, of course, but as long as the matching algorithm is reasonably sophisticated, you should still have a useful tool for flagging possible problem paintings for further investigation.I think that's right, though the problem of course is which database to check. At present there are a number of different theft databases. The largest and most successful is the Art Loss Register. However that site is not accessible to the public at large. You have to pay for and request the ALR to conduct its own search of its data. Though this technology would seem to make that process easier, Julian Radcliffe, the chairman and most vocal proponent of the ALR says "None of the imag matching is good enough to replace the art historians we use."
That may be true, but as I've argued before, the first company which figures out how to make a simple, universal and easy-to-use database will really stand out, and will also really help to legitimize the art and antiquities trade generally. Until such a database exists though, we will continue to see good faith purchasers buying stolen or illicitly excavated works leading to the classic art law dispute between an original owner and a good faith purchaser. In these cases both parties are relative innocents and the law can have a difficult time evaluating the respective claims.