Sign in to view source links and access this dataset
Description
509 pairs of Sentinel-2 satellite images and corresponding flood segmentation masks, split into training, validation, and test sets. The dataset was released by authors E. Portalés-Julià, G. Mateo-García, C. Purcell, and L. Gómez-Chova in 2023 and is hosted on Hugging Face. It requires approximately 76GB of disk space.
Use Cases
Training semantic segmentation models for flood detection based on Sentinel-2 imagery.
Benchmarking flood mapping algorithms using the provided train/val/test splits.
Studying global flood patterns and disaster response based on satellite-derived flood masks.
Strengths
Contains 509 image-mask pairs, providing a substantial corpus for model training.
Includes a predefined split into train, validation, and test sets for standardized evaluation.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
Row count is unknown, which may limit suitability assessment.
The description metadata is limited; actual data quality requires manual inspection after download.
Provenance
Source
Hugging Face dataset repository 'isp-uv-es/WorldFloodsv2'.
Collection Method
Pairs of Sentinel-2 satellite images and manually or algorithmically generated flood segmentation masks.
Freshness
Last updated 2025-07-31 16:06:52; freshness should be verified.
Geography
Global coverage is suggested by the title and publication context.
The dataset requires approximately 76GB of hard-disk space.