This project aims to process high resolution 4-band satellite images (.tif images) and classify various regions into 8 separate classes viz., Road, Tree, Bare Soil, Rail, Building, Field, Water and SwimmingPool. There are unclassified pixels present as well.
The architecture used is a modification of the existing UNet model (We have clubbed UNet with a block of CloudNet).
Our training dataset consists of 13 .tif satellite images with their corresponding ground truths and similarly we have 1 .tif cross validation image.
pillow matplotlib libtiff numpy scipy==1.1.0 pandas sympy glob2 glob3 opencv-python scikit-learn scikit-image keras tensorflow == 1.15.2