Building deep learning applications to combat Schistosomiasis

[updates: with our paper got accepted, this repo is going to be updated by the end of March, 2022]

The main contribution of this repo is schisto_model.py, which builds and runs a u-net for recognizing snail habitat in satellite images of Senegal. The script should contain sufficient information on how to run it in the pydoc.

There are a few other scripts in this directory:

convert_image.py turns an 8 channel satellite image into a 3 channel RGB image. It can also display the image.

split_mask.py turns a single mask with 4 values for the outputs into 4 separate mask files of the format expected by schisto_model.py.

straighten_rectangle.py extracts quadrilaterals from images in cases where valid input does not fill the entire rectangle.

Images are read using rasterio, which may require some installation: https://rasterio.readthedocs.io/en/stable/installation.html