Inspired from GDL/RNCan( https://github.com/Abdielfer/geo-deep-learning) wroks, DL4Flood intent to create an environment with main tool to apply deep learnind technique to flood modeling.
The data is a set of path estracted with specific size, overlap and minimum annnotation percent. preparing steps: spilt trainng set in training and validation (default trn(.9); val(.1)) make permanent transformation on trn set to encrease the number of images and the possibilities for the algorithm to see while training.