FRED: First Multi-Modal Dataset for Flooded Road Autonomous Driving
The Flooded Road Environments Dataset (FRED) represents the inaugural multi-modal dataset for autonomous driving that focuses on water-related hazards on roadways. It comprises images taken with a 2.3 MP FLIR Blackfly USB3 camera, point clouds generated by a 64-beam Ouster OS1-64 LiDAR, and IMU data from an iXblue ATLANS-C, corrected using Geoflex RTK GNSS. Data collection occurred at five sites during and following flooding incidents. The dataset is available in both KITTI-style and RTMaps formats, complete with semantic labels for identifying water hazards. Additionally, it includes position, velocity, and data from dry conditions to aid in location-based detection techniques.
Key facts
- FRED is the first multi-modal autonomous driving dataset for flooded roads.
- Dataset includes 2.3 MP FLIR Blackfly USB3 camera images.
- LiDAR data from Ouster OS1-64 (64-beam).
- IMU data from iXblue ATLANS-C corrected by Geoflex RTK GNSS.
- Data captured at five separate locations during and after flooding.
- Released in KITTI-style and RTMaps formats.
- Semantic labels provided for water hazard detection.
- Includes position, velocity, and dry-condition data.
Entities
—