A quick review of TinEye
TinEye is an image-to-image search engine from Idée. It is in a closed testing but I got to try it a couple of days ago. After a very positive review at TechCrunch, I decided to write up my impressions (a review of an earlier version is here).
They don’t make wild claims about being able to do face identification or similar (unsolved) problems. The goal seems very simple: find copies of images. With this task TinEye does a fairly good job. It finds even ones that have been modified – noise, color, stretch, crop, some photoshopping. It does not do well with rotation. That’s a major drawback (compare to Lincoln from MS Research).
These are the images that I tried.
Barbara: found both color and bw copies and a slightly cropped version.
Marilyn: found cropped and stretched versions, and an even edited (defaced) version.
Lenna: found both color and bw, but not partial or rotated versions (even though a rotated version is in the index).
Peter,
Thanks for giving TinEye a try.
A small correction: our image identification algorithms can handle rotation exceptionally well. We have not rolled that out in TinEye because it would add some computational overhead and is not a critical feature for most of our users. I hope you realize that the current TinEye index is 487 million images and that we are working on growing that into the tens of billions of images. While it would be a nice feature, rotation invariance is not as significant as many of the goals and features we have in the works for TinEye. Thanks for the link to Lincoln from MS Research but I have never seen it work on a large scale collection (1/2 a billion images would be a good start!).
Keep up the good work!
Cheers,
Leila
I can see that for TinEye’s purposes rotation isn’t very important. But it would definitely not hurt. I care about this because my interest is mostly image analysis and search in science applications. BTW, in case it didn’t come across in the post, my impression of TinEye is quite positive. It works!