A team led by Antonio Torralba, assistant professor in MIT’s Computer Science and Artificial Intelligence Laboratory has been trying to find out the shortest numerical representation the smallest amount of information, necessary to indicate the contents of an image. This will be useful in categorizing and searching billions of images on the internet.

Today, the only way search engines can search images is through the captions or titles people add to their images. There are no other ways. Sometimes these captions are misleading, and there are pictures with no captions at all which makes it harder to find them. When CPUs are able to see images this will help search engines a lot. It will also help people when they download their pictures from their cameras to their PCs, the PC will automatically be able to identify the pictures’ contents and caption them. It won’t be necessary for people to put captions to pictures one by one.

“We’re trying to find very short codes for images,” says Torralba, “so that if two images have a similar sequence [of numbers], they are probably similar–composed of roughly the same object, in roughly the same configuration. With very large amounts of images, even relatively simple algorithms are able to perform fairly well”.

Source: sciencedaily.com