Datasets

Viona on narrow forest road (ROS bag file, 29.2GiB)
Supplementary files (calibration and readme)

Our robot Viona being driven on a very narrow forest road on the mountain Schauinsland close to Freiburg, Germany.
The logged sensors include the robot's odometry, a Velodyne HDL 64, four Bumblebee stereo cameras and an Applanix navigation system.

Autonomous run on Schauinsland (ROS bag file, 36GiB)
Supplementary files (calibration and readme)

Our robot Viona is going fully autonomously through an outer urban environment on the mountain Schauinsland close to Freiburg, Germany. The only prior knowledge of the area the robot has is an OpenStreetmap road network on which we mark goal points manually. The robot then uses its internal sensors to classify the type of environment it is moving on and detects obstacles that it has to circumvent. The classification of the road is used to correct the positions of the goal points coming from the OpenStreetMap road network.
The logged sensors include the robot's odometry, a Velodyne HDL 64, four Bumblebee stereo cameras and an Applanix navigation system.

Autonomous run through Mooswald (ROS bag file, 41GiB)
Supplementary files (calibration and readme)

Our robot Viona is going fully autonomously through an outer urban environment at the Mooswald forest in Freiburg, Germany. The only prior knowledge of the area the robot has is an OpenStreetmap road network on which we mark goal points manually. The robot then uses its internal sensors to classify the type of environment it is moving on and detects obstacles that it has to circumvent. The classification of the road is used to correct the positions of the goal points coming from the OpenStreetMap road network.
The logged sensors include the robot's odometry, a Velodyne HDL 64, four Bumblebee stereo cameras and an Applanix navigation system.

Body Part Segmentation Dataset

We introduced two new robotics datasets for human semantic part segmentation: Freiburg Sitting People and Freiburg People in Disaster. They provide high resolution data for experiments on ground and aerial robot segmentation applications.
Deepscene Classification

Online demo and dataset for camera-based semantic terrain classification using convolutional neural networks.