Visualized in Foxglove: The Rellis-3D UGV Dataset
Try it for yourself: https://foxglove.dev/examples 🔥
Semantic scene understanding is crucial for autonomous navigation, particularly so in off-road environments.
The Rellis-3D dataset contains high-resolution lidar and image data with annotations for semantic segmentation, covering challenging environments that typical urban datasets do not encompass, such as varied terrain types under different weather conditions. This also presents challenges to existing algorithms related to class imbalance and environmental topography.
The dataset features multiple sensors, including a 64-beam Ouster lidar, a 32-beam Velodyne lidar, a Nerian Karmin2 Stereo Depth Camera (without RGB), a Nerian Stereo Depth Camera (without RGB), and a Vectornav VN-300 Dual Antenna GNSS/INS (Inertial Navigation System with GPS/IMU).
The diverse sensor suite and challenging environments captured in the dataset make it valuable for off-road robotics and autonomous vehicle navigation research.
The Rellis-3D dataset has been provided by the research of Peng Jiang, Phil Osteen, Maggie Wigness, and Srikanth Saripalli from the Unmanned Systems Lab at Texas A&M University and US Army DEVCOM Research Lab.
Link to the dataset and to learn more about the project:
https://github.com/unmannedlab/RELLIS...
https://www.unmannedlab.org/research/...
https://clearpathrobotics.com/warthog...