Dataset for fluvial sediment segmentation from remote sensing imagery using U-Net and incremental learning (Amazon case study)

This repository contains the data and supporting materials used to develop mlfluv, a deep learning framework for fluvial sediment segmentation from remote sensing imagery. The model is based on a U-Net architecture and uses an incremental learning strategy to improve the detection of fluvial sediment, a minority and spatially heterogeneous land cover class that is often confused with other bare surfaces in conventional land cover products. The dataset was prepared for an Amazon case study and is intended to support model training, validation, testing, and reproducibility of the reported results. Detailed descriptions of the dataset construction, label sources, model design, and experimental workflow are provided in the manuscript “Identifying fluvial sediment using remote sensing data and deep learning: an Amazon case study”.