Compile Sentinel output in 3-D matrix, output to file (geotiff format) and store locally for input to the CNN. This may require re-aquisition of images in location where we can obtain ~500 images (buffer for clouds, missing data, etc.); we need about 200 to train.
Learn convolutional neural networks (CNN):
https://www.datacamp.com/tutorial/neural-network-models-r
Compile Sentinel output in 3-D matrix, output to file (geotiff format) and store locally for input to the CNN. This may require re-aquisition of images in location where we can obtain ~500 images (buffer for clouds, missing data, etc.); we need about 200 to train.