UrbanFloodCastV1

UrbanFloodCastV1 is the first version repository for Inno_Maus project. This repository contains the necessary scripts for data processing, model training, and the required dependencies to run the project.

Table of Contents

• Overview
• Data pixelization
• Model Training
• Installation
• Usage

Overview

UrbanFloodCastV1 leverages deep learning techniques to forecast urban flooding based on hydrology data. The project uses a variety of libraries and tools for data preprocessing, model building, and visualization.

Data pixelization

The pixelized data required for model training and evaluation can be downloaded from the following link:

Pixelized Data

To get this data, you can process the data using the script provided in data/process.ipynb. This can help convert .txt to .tif files.

Model Training

We prepared the pre-trained model checkpoints: Checkpoints

Installation

To set up the project locally, follow these steps:

1. Clone the repository:

   git clone https://github.com/zhu-xlab/UrbanFloodCastV1.git
   cd UrbanFloodCastV1
   

2. Install the required dependencies: You can install all the required libraries using the requirements.txt file::

   pip install -r requirements.txt
   

Usage

To use the model, execute the following steps:

1. Process the runoff / ground truth data by running the downsample.ipynb notebook. This can help downsample the images. Downsampled runoff is here.

2. Please change the 'Root_path' to your local path.

3. Put the test ground truth events folder to 'data/val'. Test the model using the load_model.ipynb notebook.