the HybridSAR Road Dataset (HSRD)
The HybridSAR Road Dataset (HSRD) is a comprehensive dataset designed to facilitate road segmentation tasks using Synthetic Aperture Radar (SAR) imagery. This dataset addresses the scarcity of high-resolution annotated SAR data by integrating real and synthesized SAR datasets. HSRD includes sub-meter resolution SAR images and corresponding road labels, making it an invaluable resource for researchers and practitioners in remote sensing, urban planning, and disaster management.
The HSRD dataset comprises three main components:
SpaceNet 6 Road (SN6R) Dataset: This dataset includes high-resolution SAR images and road labels sourced from OpenStreetMap (OSM). The SAR images are derived from the SpaceNet 6 Challenge, covering the city of Rotterdam.
DG-SAR Dataset: Synthesized from the DeepGlobe (DG) optical dataset using an EO2SAR translation model. This dataset provides SAR-like images generated from high-resolution electro-optical (EO) images.
SN3-SAR Dataset: Synthesized from the SpaceNet 3 (SN3) optical dataset using an EO2SAR translation model. Similar to DG-SAR, this dataset offers SAR-like images generated from high-resolution EO images.
Download the dataset here.
This dataset is released under the MIT License.
We would like to thank the SpaceNet team, DeepGlobe Challenge organizers, and OpenStreetMap contributors for providing the foundational data used to create this dataset.
For any questions or issues, please contact your.email@example.com.