JamUNet: predicting the morphological changes of braided sand-bed rivers with deep learning

This repository stores the data, code, and other files necessary for the completion of the Master's thesis of Antonio Magherini, student of the MSc Civil Engineering program - Hydraulic Engineering track, with a specialisation in River Engineering at the Faculty of Civil Engineering and Geosciences of Delft University of Technology (TU Delft). The thesis was carried out in collaboration with Deltares. The manuscript can be found at TU Delft repository. For any information, feel free to contact the author at: [email protected]. The image represents the Brahmaputra-Jamuna River at the border between India and Bangladesh. The image was taken on January 25, 1994. It was retrieved from Google Earth Engine USGS Landsat 5 collection.

---

Repository structure

The structure of this repository is the following:

• benchmarks, contains modules and notebooks of the benhcmark models used for comparison;
• data, contains raw data (satellite images, river variables);
• images, contains the images shown in the thesis report and other documents; (to be added soon)
• model, contains the modules and noteboooks with the deep-learning model;
• other, contains documents, images, and other files used during the project; (to be added soon)
• postprocessing, contains the modules used for the data postprocessing;
• preliminary, contains the notebooks with the preliminary data analysis, satellite image visualization, preprocessing steps, and other examples;
• preprocessing, contains the modules used for the data preprocessing.

---

Install dependencies

braided.yml is the environment file with all dependencies, needed to run the notebooks.

To activate the environment follow these steps:

• make sure to have the file braided.yml in your system (for Windows users, store it in C:\Windows\System32);
• open the anaconda prompt;
• run conda env create -f braided.yml;
• verify that the environment is correctly installed by running conda env list and checking the environment exists;
• activate the environment by running conda activate braided;
• deactivate the environment by running conda deactivate;

---

Cite

Please cite the Master thesis as:

@mastersthesis{magherini2024,
author = {Magherini, A.},
title = {{JamUNet: predicting the morphological changes of braided sand-bed rivers with deep learning}},
school = {{Delft University of Technology}},
year = {2024},
month = {10},
howpublished = {\url{https://repository.tudelft.nl/record/uuid:38ea0798-dd3d-4be2-b937-b80621957348}}
}