The visual-based search technology meets the challenge of "a million or none" in image search.Media Bakery LLC this week announced that it has implemented VIMA Technologies Inc.'s MindReader search technology platform to bring visual searching capability to its stock photo site. The added search capability, which the company has been beta testing for several months, will officially debut on Media Bakery's site on August 8.
Chad Newell, CEO of the Carpinteria, Calif.-based company said Media Bakery chose VIMA's technology because it provided a better way for graphic and creative professionals to find images quickly in Media Bakery's collection of more than 750,000 images and bridged what he called a language barrier in image search.
"We carry many collections from worldwide suppliers and one of the limitations on relying just on keyword search is the language barrier that exists in the descriptive adjectives used to caption images. We know this visual search technology addresses that barrier," Newell said.
MindReader visual search works by tapping into all the keyword searching strengths that exist on a site and adding to them the benefit of searching by the visual aspects of the image as well, according to David Telleen-Lawton, CEO of Goleta, Calif.-based VIMA.
Lawton said MindReader solves "the million or none" challenge.
"When agencies or creatives are looking at any large data shed of images, what they have to combat with keyword search is depending on the keywords to bring forth the images they want. Typically in this arena what happens using just keywords is you'll get a million images after the initial keyword search, and then as you continue to type more specific keywords, at some point you'll get no images coming up," Lawton said.
"And as opposed to other technologies for visual search that have a single click, [with the] 'More Like This' [feature,] which brings up the best matches based on that one image, MindReader allows the searcher to continue refining with positive and negative feedback. The searcher keeps selecting good and bad examples and the result is that the top most images get more and more like their perfect image. So it is the multiple clicks and the benefit of negative feedback that differentiates MindReader," he said.
According to Lawton, MindReader uses many images to form a model file of what the searcher is seeking and uses that to reorder the images. MindReader does not eliminate images but instead continuously reorders them so the best image or images are at the top.
"The power of combining many images to form a composite search model file is unique to MindReader. It boosts searches by allowing the searcher to bring the best matches to the top rather than having to go look for the images. It is also possible to start a search by uploading an image at the beginning too," Lawton said.
Another stock photography company employing the search technology with marked success is the Danita Delimont Stock Photography agency.
The agency's founder and CEO, Danita Delimont, former national president of the American Society of Picture Professionals said, "I've seen how this technology has evolved in the past five years and this really is cutting-edge technology that lets me compete right up there with the big boys.
"It helps our users significantly reduce their search time by helping them continually refocus their image search," Delimont said.
Delimont is committed to letting her users take advantage of the technology and created and produced an educational video based on how to use the image-based search solution for her agency's Web site.
Delimont predicts that long term, the only challenge for MindReader and agencies that employ it will come in the form of engaging image researchers to try it out and take the time to learn how to use it.
