Search:

Home | Root Category | Internet

Start Trading

Google's New VisualRank the Best Image Search?

By: .JohnSmith.

Google can't stop innovating, again. It's their lifeblood to continually develop and deliver the internet's hottest search technologies. We've come to expect this out of the world's internet search giant. A new search system, named VisualRank, will liken an image search to that of a Google textual search, suggests a paper written by Google scientists Yushi Jing and Shumeet Baluja.

VisualRank will allow web surfers to search for images and return results with much more precision and accuracy compared to Google's current image search, which is more of a hit or miss process.

The differences between the new image search and the old image search are that the new search partially relies upon "incorporation of visual signals in this [search] process." What this means, using McDonald's as a colorful example, is that VisualRank will be able to recognize common "visual themes" within a group of images - basically common trends within millions of images. When somebody conducts a search for "McDonald's" using the new technology, the system will be able to recognize that the image of the golden arch - commonly considered the most popularized image of McDonald's - is the most relevant image to the search query. The system therefore prioritizes images with a golden arch into the search results, therefore increasing the relevancy of the returned results. The image below shows the relevancy web a a search for an image of the Mona Lisa (image in center).

In relation to the old system, Google's VisualRank will no longer only search for image queries based off textual clues alone.

Tests of this new VisualRank system returned 83 percent fewer irrelevant search results than Google Image Search, according to the VisualRank paper, leading enthusiasts to believe that such a system could completely reinvent people's ability to search for specific images and return extremely relevant results.

There's a couple promising implications if this technology really works as well as Google claims. Imagine being able to conduct a search for a product based off a set of keywords, and receive ranked returned results of products that most similarly look like your initial query. For example, if you search for a specific tee shirt that has a specific design on it, you'd get returned results of that tee shirt and other tee shirts that resemble the product you originally searched for. VisualRank could thus become an important referral or recommendation device to better the online shopping experience.

Like.com servers as the biggest competitor to VisualRank, who product enables you to search shopping inventories based off visual recognition, returning similar-looking results. On Like.com, you can further select between the type of visual search you want, being able to pick between detail search, color match, shape match, or pattern match. A person could potentially highlight a an image of a watch and conduct detail search that would return similar looking watches.

I decided to put Like.com's abilities to the test. I conducted a detail search on the shirt below to the left for any other pieces of clothing that resembled or included the Miami Dolphins logo, seen in the center of the shirt (the box designates the area that I'm talking about).

The first result returned included a tee shirt that resembled nothing like the Miami dolphin logo. The images that come to my mind when thinking about dolphins for some reason does not include race cars, although I suppose people are entitled to their own opinions. There's a possibility that Like.com simply didn't have anything in their inventory that looked like a Miami Dolphin, but I think I'd prefer the honest answer rather than crazy race car art.

Either way, visual search technologies undoubtedly have a future in many different markets: online shopping, sciences, arts, etc. Further research and development may increase these technologies abilities. It'll be interesting to see how Google's VisualRank compares when it goes live.

Article Source: http://www.associatedigest.com

For more articles about internet business, technology, entrepreneurship, and startups from this author, log onto www.gothamtechminute.blogspot.com

Please Rate this Article

 

Not yet Rated

Click the XML Icon Above to Receive Internet Articles Via RSS!

Powered by Article Dashboard