The Semantic Robot Challenge is a photo treasure hunt for robots. Competing mobile robot platforms equipped with multiple cameras and laser range scanners are given a text list of objects dispersed on furniture or on the floor in a conference room. They can learn about what the objects are supposed to look like by connecting to the Internet and finding images of them. Then they have to autonomously search the environment and take snapshots of as many objects as they can, adding bounding boxes around the objects in the pictures and labeling them. The winner is the robot that takes pictures of the largest number of objects, with correct bounding boxes and labels (evaluated by human judges according to well-defined rules). This competition challenges the teams to successfully implement a large number of strategies that humans would use to accomplish the same task. Successful robots are able to perform navigation, exploration, obstacle avoidance and mapping, peripheral vision, foveal vision, observation from multiple viewpoints, bottom-up feature detection and top-down object model verification.
The PI is with the organizing committee of the Third Semantic Robot Vision Challenge (SRVC), and this NSF award supports travel scholarships for all participating teams. This challenge is scheduled to take place as a Special Track of the Fifth International Symposium on Visual Computing (ISVC09) in Las Vegas on December 1, 2009. The first SRVC event took place at the AAAI Conference in Vancouver, British Columbia, Canada in 2007. The second SRVC event was organized at CVPR in Anchorage, Alaska in 2008. The practice of providing travel support to participating teams has been shown to provide a great encouragement for teams to attend, and it was implemented for the past two SRVC events.
The web site for the Semantic Robot Vision Challenge is www.semantic-robot-vision-challenge.org/
