Visual Search That Actually Works: TinEye
A new visual search engine has been launched by a Canadian company TinEye.
Instead of entering text describing an object such as the phrase “tiger jumping” you upload or provide the URL of a photo.
But this is far more than a simple file search. I tried with jpg and gif photos and it found both file types on the internet regardless of which I provided. It even found photos where they had been cropped or the colour and size changed.
Passing it a collage of several photos it was able to identify other photos containing one of the component photos.
Clearly it is performing a rapid search in its database for some kind of signature of the visual look of the image and this is allowing a range of images containing that pattern to be identified.
It would seem to be mainly useful for researching the origins and usage of a photo in order to obtain, or identify abuse of, the usage rights.
I was very impressed with the tests I performed.
If you would like to give it a go it is at http://tineye.com/
I would be intereseted in hearing where people find the limits are. I ran out of good tests without hitting a problem.
Recent Comments