A dataset for tree-crops prediction in Sentinel-2 imagery
The dataset was designed for semantic-segmentation-based deep learning models and utilized in the AI4Copernicus service: “Deep network for pixel-level classification of S2 patches" in which a U-net neural network was used in order to create an AI model for tree-crops prediction.
The dataset and the U-net model was created for the SandMap platform, a tangible GIS educational platform which aims to inform the students and citizens about the importance of precision agriculture and the relative cutting-edge technology in order to automate significant processes such as classification and localization of crops using AI. Thus with the use of:
-- the AI4Copernicus service: “Deep network for pixel-level classification of S2 patches"
-- the designed dataset which includes pairs of original and mask image patches by Sentinel-2 satellite
The user is able to train a U-net model with ResNet50 as a backbone in order to create a model for tree-crops prediction.