This is Tensorflow based project using U-Net Architecture to map flooding using Sentinel-1 imagery.

You can download our dataset at this link https://www.kaggle.com/dimskeee/unet-segmentation-using-radar-imagery, which includes thousands of tiles that are used to train, test, and validation. The size of the tiles is 128 x 128, and the default filter in the script is 16, 32, 48, and 64.
Folder Structure:

Images structure:

Note: Training for this model was done on an RTX 2060 Super 6GB