FloodPlanet

FloodPlanet is a curated dataset specifically designed for flood detection research, making it an invaluable resource for the geospatial and machine learning communities. It comprises manually labeled inundation data, providing high-quality ground truth for developing and testing advanced algorithms. The dataset is particularly geared towards deep learning applications, enabling researchers to train robust models for identifying flooded areas.

This dataset leverages a diverse range of satellite imagery sources, including PlanetScope, Sentinel-1, and Sentinel-2. This multi-sensor approach enhances the dataset's utility, allowing for comprehensive analysis of flood events across various environmental conditions. Key subjects addressed include flood detection, manual labeling techniques, and the application of deep learning methodologies.

The FloodPlanet dataset, version 1.0, was primarily contributed by Zhijie(JJ) Zhang, alongside Rohit Mukherjee, Andrew Molthan, Alexander Melancon, Beth Tellman, and Jonathan Giezendanner. It is hosted within the CyVerse Data Commons, featuring a structured organization that includes the main FloodPlanet data, a STAC catalog for easy discovery and access, and a README file for detailed information. This resource facilitates open research into global flood monitoring and response.