The progress made in training algorithms to recognize images via the ImageNet Large Scale Visual Recognition Challenge has prompted its 2018 replacement with the more formidable task of teaching robots to perceive the world in three dimensions (3D).
Compiling a database of images that includes 3D information would enable robots to be trained to identify objects around them and plot out the best navigation routes, says Victor Prisacariu at the University of Oxford in the U.K.
The University of North Carolina at Chapel Hill's Alex Berg says the database for the new contest would be comprised of digital models of real-world environments or 360-degree photos that include depth information. He expects it will be at least five years before robots consistently understand their surroundings and can explain what they see as well as humans.
Imperial College London's Andrew Davison notes the new competition also should further virtual and augmented reality technologies.
From New Scientist
View Full Article
Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA
No entries found