RICA: Robocentric Indoor Crowd Analysis Dataset

In this paper, we introduce an egocentric dataset recorded from a robot’s point of view (robocentric), which has been created to serve as a platform for indoor crowd analysis. The dataset features over 100,000 RGB, depth, and wide-angle camera images as well as LIDAR readings, recorded during a social gathering where the robot captured group interactions between participants using its on-board sensors. We evaluated three different human detection algorithms on our dataset to demonstrate the challenges of indoor crowd analysis from a robot’s perspective.