|intended to do for digital images .. what .. PageRank software did for searches of Web pages |
It's not surprising Google's managed to
whip up an "image recognition" and
I'm betting they roll out an improved version
of the "indexing the spoken audio in videos"
systems that are now beginning to appear,
very shortly, as well.
Therefore, one could do an image search on a supermodel such as Christie Brinkley or Tyra Banks and come up with photos of women who look like them?
This is going to be annoying for me, because I look just like George Clooney.
ya, I see a lot of:
instances in my log files coming soon! =/
"ranking images that look most similar."
PageRank has got (almost) nothing to do with similiarity which is a completely different problem altogether (even though PR is certainly used to pick the highest match from those found to be similar).
|concentrated on the 2000 most popular product queries |
|With a team of 150 Google employees, it created a scoring system for image “relevance.” |
|The researchers said the retrieval returned 83 percent less irrelevant images. |
I don't understand how 150 employees scoring a set of 2000 images to reduce irrelvance by 83% = software technology
Humans can provide scoring which is analysed for patterns by software and applied to new, unscored by humans, images. Still, 150 people in that team sounds rather high.
It's simple technology really.
You know how on Flickr people draw boxes around parts of the image? Or facebook lets you tag people's names?
Now if you search for "boobie", you'll get square snippets of boobies cropped out of flickr images of friends and family.
Or if you search for "Marsha", they'll crop out Marsha from the facebook images. Brilliant.
Im not impressed, but another thing is I think we will soon see something in the image rankings, be cause normaly we see a update in images rankings every 3 month, but that has past some time ago now.
This is perfect! Now all I need is a mashup that takes Yahoo Local Circles and Google VisualRank so I can find a Charlize Theron look alike within 50 meters! Awesome :)
Sounds like Like.com I'm sure Burak Gokturk will be interested in Google's "new" image search.
|Burak Gokturk -he developed computer vision algorithms at Canesta,Intel and BEKO. He has been in the vision field for more than 8 years,published more than 30 papers, and holds more than 15 patent applications. His computer vision algorithms have been prototyped and/or implemented in various computer vision products. He developed a 3D shape representation scheme, called "Random Orthogonal Shape Sections", which achieved one of the best recognition rates for the detection of colon cancer. His other research concentrated on tracking, detection and recognition of 3D shapes and faces, medical imaging and 3D image compression. |
this sounds good for sure and might lead to better opportunity for smart developers.