Flooded Road Environments Dataset (FRED) is an autonomous vehicle dataset for detecting flooded roads during on-road deployment. It was collected using a modified Renault Zoe equipped with front and rear cameras, a LiDAR sensor, and a GNSS-corrected IMU. Data was gathered from 5 separate locations around Brisbane, Australia, both during and after flood events.
Use Cases
- Train computer vision models to detect flooded roads based on camera imagery.
- Develop sensor fusion algorithms using combined camera, LiDAR, and IMU data.
- Test autonomous vehicle robustness in adverse weather conditions based on real-world sensor data.
- Benchmark object detection and segmentation models on a domain-specific environmental hazard.
Strengths
- Data collected from 5 distinct geographic locations, providing environmental variety.
- Sensor suite includes multiple modalities: front and rear cameras, a LiDAR, and a GNSS-corrected IMU.
- Data collection targeted a specific, challenging real-world condition: flooded roads.
Limitations
- Description metadata is limited; actual data quality requires manual inspection after download.
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
Provenance
- Source
- Author CMalone-Jupiter on Hugging Face.
- Collection Method
- Collected using a custom-modified Renault Zoe autonomous vehicle sensor stack.
- Freshness
- Last updated 2026-05-22 03:16:54; freshness should be verified.
- Geography
- Brisbane, Australia.