NIST and Willow Garage: Solutions in Perception Challenge
Willow Garage and the National Institute of Standards (NIST) are launching a new type of robot perception challenge aimed at both:
- establishing which perception problems have solutions and
- advancing perception solutions to new domains.
Figure: A 3D textured point cloud of some of the objects. In this image 4 of the 5 objects are recognized with coordinate systems overlaid representing the object pose relative to the robot.
Most existing challenges in computer vision and perception aim at making progress in complex object class identification tasks such as pursued by the PASCAL Visual Object Classes challenge. These contests are useful, for example, in creating progress in web image retrieval tasks. In contrast, the Solutions for Perception Challenge aims at advancing sensing based robotics by identifying domains where roboticists can rely on having a working perception solution. Another key difference from other perception challenges is that to be of use for robotics, a perception solution must not only identify objects but calculate the 3D pose of the object relative the the robot.
The challenge will debut at the IEEE International Conference on Robotics and Automation (ICRA) 2011 in Shanghai, China, on May 9-13, 2011. There will be a somewhat unusual cash prize structure for contestants: up to $10,000 dollars will be awarded exponentially starting at $3.50 for a first prize winner who achieves 80% recognition and increasing exponentially from there. This prize structure reflects the increasing difficulty of achieving scores closer to 100%. The first contest focuses on recognizing rigid textured objects.
Joining the contest
This challenge is organized by Gary Bradski (Willow Garage) and Tsai Hong (NIST). Further encouragement and support has come from Thomas Kalil, Associate Director for Policy at the White House Office of Science and Technology and from Elena Messina, Chief of Intelligent Systems Division at NIST.