Deep Learning Model Training and Validation Data for Global Floating Algae Detection

This dataset contains the training and validation data used to develop and evaluate a ResUNet deep learning (DL) segmentation model for detecting floating algae from MODIS/Aqua imagery at the global scale. The model was trained using inputs including MODIS Rayleigh corrected reflectance (Rrc) in 7 spectral bands and the Alternative Floating Algae Index (AFAI), and is capable of identifying both microalgae (phytoplankton) scums and macroalgae (seaweed) mats. These include Trichodesmium, Noctiluca, Dinoflagellates, Cyanobacteria, Sargassum, and Ulva.