A Deep Learning Based Method to Delineate the Wet/Dry Shoreline and Compute its Elevation Using High-Resolution UAS Imagery

Content

This repository contains the code used in the described paper

Publication

Publication under review

Citation

Vicens-Miquel, Marina; Medrano, Antonio; Tissot, Philippe; Kamangir, Hamid; Starek, Michael. A Deep Learning Based Method to Delineate the Wet/Dry Shoreline and Compute its Elevation Using High-Resolution UAS Imagery. Remote Sensing MDPI. (Submitted, pending review)

LaTeX Citation

    @article{vicensmiquel,
            title={A Deep Learning Based Method to Delineate the Wet/Dry Shoreline and Compute its Elevation Using High-Resolution UAS Imagery},
            author={Vicens-Miquel, Marina and Medrano, Antonio and Tissot, Philippe and Kamangir, Hamid and Starek, Michael},
            journal={Remote Sensing MDPI(Submitted, pending review)}
    }

Installation

1. Install miniconda
       wget https://docs.conda.io/en/latest/miniconda.html
       ./Miniconda3-latest-Linux-x86_64.sh
   
2. Install mamba
       conda install mamba -n base -c conda-forge
   
3. Create an environment
       conda create --name tf_gpu
   
4. Activate the environment
       conda activate tf_gpu
   
5. Install TensorFlow
       mamba install tensorflow-gpu -c conda-forge
   
6. Install packages using pip
       keras == 2.6.0
       matplotlib == 3.5.1
       pandas == 1.4.1
       numpy == 1.22.0
       scikit-learn == 1.0.2
   

Quickstart

To train the model, please make sure to first activate the environment. After, please type the following command to start training python src/train.py

Data Format

The dataset is organized under the below directories. testingOrtho1 is the name of the dataset used. In this case, testingOrtho1 means that we are using the orthomosaic 1 as an independent testing dataset, while using all the other orthomosaic for training and validation. Depending on which orthomosaic we want as the independent testing dataset, a different dataset will be loaded.

data
└── testingOrtho1
    ├── training
    │   ├── original (this directory contains all the original images) 
    │   └── labeled  (this directory contains all the labeled images)
    ├── testing
    │   ├── original (this directory contains all the original images) 
    │   └── labeled  (this directory contains all the labeled images)
    └── validation
        ├── original (this directory contains all the original images) 
        └── labeled  (this directory contains all the labeled images)