Help teach robots to see - with your Kinect

Visual Dictionary and Visual Browsing revisited by Devin Coldewey

EXTRACT:

Swedish researchers are hoping to create a visual dictionary to improve the ability of robots to understand the world around them, but they can't do it alone. They need your help — if you've got a Kinect.

Robots are smart in that they can store tons of information and process it quickly, but unlike humans they haven't been walking the Earth for long, and have very little practical knowledge of common objects. How heavy is a coffee mug? No data, though a human could easily size it up and make a good estimate. What's that thing on the floor? A human would recognize it as a shoe, but to a robot, it could be anything.

A project called Kinect@Home is being led by Alper Aydemir at Sweden's Royal Institute of Technology. Aydemir hopes that by outsourcing that common knowledge to humans, robots will develop a better idea of their surroundings. To that end, he has started a database of 3-D models of everyday objects like books and shoes, captured with the cheap and effective Kinect. continue reading