CrowdSurfer

The CrowdSurfer dataset consists of 56 ROS2 .bag files captured in densely populated environments. It includes two primary subsets: the ICRA subset with 37 .bag files and the IROS subset with 19 .bag files. Each file captures real-world robotic navigation scenarios involving dense human crowds. The data comprises synchronized sensor streams such as RGB-D images, LiDAR scans, IMU readings, and odometry, suitable for perception, localization, and social interaction modeling. The dataset reflects both indoor and semi-structured public spaces, enabling robust testing of navigation and prediction algorithms in human-centric scenes.