Case study reviewing feasibility of training a U-Net based Convolutional Neural Network model to automatically identify and delineate areas of qat agriculture in Sentinel-2 multispectral imagery acquired over central Yemen. Source code developed for this study leverages functionality of the unet-keras-collection library.
Methodology and initial results are presented in a Jupyter notebook.