Collaborative online game trains computer vision algorithms

New Scientist reports on Peekaboom, the fun collaborative online game where you help train computer vision recognition algorithms. The aim is to identify an image when only a portion of it is visible. From the article:

Currently, computers or robots apply statistical rules to a database of known images to identify new ones. Although relatively effective, this requires thousands of images and hours of training by humans. Peekaboom aims to ease the burden by harnessing the brain power of willing web users.

"The collected data can be applied towards constructing computer vision algorithms, which require massive amounts of training and testing data not currently available," von Ahn told New Scientist. "The target database will contain millions of images, all fully annotated with information about what objects are in the image, where each object is located, and how much of the image is necessary to recognise it."