According to the official Semantic Robot Vision Challenge website "The Semantic Robot Vision Challenge seeks to fuse the state of the art research from various vision and mobile robotics communities". In a nutshell, the robot is required to recognize objects while navigating in an small room. The objects that need to be recognized are given in textual form (an electronic text file) in advance.

In the first phase of the competition the robot is given limited time to access the internet and build descriptors for these objects. Then, the robot have limited time to explore a small room filled with some of the objects on the lists as well as other objects and recognize them.

While there are similarities with the scavenger hunt competitions that took place in the previous years there is a fundamental difference. In this challenge the appearance of the object is not known in advance. So the robot is requested to build an appearance model of an object from a symbolic description i.e., its name. Also the robot is requested to generalize i.e. from a few images of an object found on the internet to create a model describing this specific object.