MagicBathyNet: A Multimodal Remote Sensing Dataset for Benchmarking Learning-based Bathymetry and Pixel-based Classification in Shallow Waters

MagicBathyNet is a benchmark dataset made up of image patches of Sentinel-2, SPOT-6 and aerial imagery, bathymetry in raster format and seabed classes annotations. Dataset also facilitates unsupervised learning for model pre-training in shallow coastal areas. It is developed in the context of MagicBathy project.



DOI of GitHub Repository

Package for benchmarking MagicBathyNet dataset in learning-based bathymetry and pixel-based classification.

This repository contains the code of the paper "P. Agrafiotis, Ł. Janowski, D. Skarlatos and B. Demir, "MAGICBATHYNET: A Multimodal Remote Sensing Dataset for Bathymetry Prediction and Pixel-Based Classification in Shallow Waters," IGARSS 2024 - 2024 IEEE International Geoscience and Remote Sensing Symposium, Athens, Greece, 2024, pp. 249-253, doi: 10.1109/IGARSS53475.2024.10641355."

Citation

If you find this repository useful, please consider giving a star ⭐.
If you use the code in this repository or the dataset please cite:

P. Agrafiotis, Ł. Janowski, D. Skarlatos and B. Demir, "MAGICBATHYNET: A Multimodal Remote Sensing Dataset for Bathymetry Prediction and Pixel-Based Classification in Shallow Waters," IGARSS 2024 - 2024 IEEE International Geoscience and Remote Sensing Symposium, Athens, Greece, 2024, pp. 249-253, doi: 10.1109/IGARSS53475.2024.10641355.

@INPROCEEDINGS{10641355,
  author={Agrafiotis, Panagiotis and Janowski, Łukasz and Skarlatos, Dimitrios and Demir, Begüm},
  booktitle={IGARSS 2024 - 2024 IEEE International Geoscience and Remote Sensing Symposium}, 
  title={MAGICBATHYNET: A Multimodal Remote Sensing Dataset for Bathymetry Prediction and Pixel-Based Classification in Shallow Waters}, 
  year={2024},
  volume={},
  number={},
  pages={249-253},
  doi={10.1109/IGARSS53475.2024.10641355}}

Getting started

Downloading the dataset

For downloading the dataset and a detailed explanation of it, please visit the MagicBathy Project website at https://www.magicbathy.eu/magicbathynet.html

Dataset structure

The folder structure should be as follows:

┗ 📂 magicbathynet/
  ┣ 📂 agia_napa/
  ┃ ┣ 📂 img/
  ┃ ┃ ┣ 📂 aerial/
  ┃ ┃ ┃ ┣ 📜 img_339.tif
  ┃ ┃ ┃ ┣ 📜 ...
  ┃ ┃ ┣ 📂 s2/
  ┃ ┃ ┃ ┣ 📜 img_339.tif
  ┃ ┃ ┃ ┣ 📜 ...
  ┃ ┃ ┣ 📂 spot6/
  ┃ ┃ ┃ ┣ 📜 img_339.tif
  ┃ ┃ ┃ ┣ 📜 ...
  ┃ ┣ 📂 depth/
  ┃ ┃ ┣ 📂 aerial/
  ┃ ┃ ┃ ┣ 📜 depth_339.tif
  ┃ ┃ ┃ ┣ 📜 ...
  ┃ ┃ ┣ 📂 s2/
  ┃ ┃ ┃ ┣ 📜 depth_339.tif
  ┃ ┃ ┃ ┣ 📜 ...
  ┃ ┃ ┣ 📂 spot6/
  ┃ ┃ ┃ ┣ 📜 depth_339.tif
  ┃ ┃ ┃ ┣ 📜 ...
  ┃ ┣ 📂 gts/
  ┃ ┃ ┣ 📂 aerial/
  ┃ ┃ ┃ ┣ 📜 gts_339.tif
  ┃ ┃ ┃ ┣ 📜 ...
  ┃ ┃ ┣ 📂 s2/
  ┃ ┃ ┃ ┣ 📜 gts_339.tif
  ┃ ┃ ┃ ┣ 📜 ...
  ┃ ┃ ┣ 📂 spot6/
  ┃ ┃ ┃ ┣ 📜 gts_339.tif
  ┃ ┃ ┃ ┣ 📜 ...
  ┃ ┣ 📜 [modality]_split_bathymetry.txt
  ┃ ┣ 📜 [modality]_split_pixel_class.txt
  ┃ ┣ 📜 norm_param_[modality]_an.txt
  ┃
  ┣ 📂 puck_lagoon/
  ┃ ┣ 📂 img/
  ┃ ┃ ┣ 📜 ...
  ┃ ┣ 📂 depth/
  ┃ ┃ ┣ 📜 ...
  ┃ ┣ 📂 gts/
  ┃ ┃ ┣ 📜 ...
  ┃ ┣ 📜 [modality]_split_bathymetry.txt
  ┃ ┣ 📜 [modality]_split_pixel_class.txt
  ┃ ┣ 📜 norm_param_[modality]_pl.txt

The mapping between RGB color values and classes is:

For the Agia Napa area:
0 : (0, 128, 0),   #seagrass
1 : (0, 0, 255),   #rock
2 : (255, 0, 0),   #macroalgae
3 : (255, 128, 0), #sand
4 : (0, 0, 0)}     #Undefined (black)

For the Puck Lagoon area:
0 : (255, 128, 0), #sand
1 : (0, 128, 0) ,  #eelgrass/pondweed
2 : (0, 0, 0)}     #Undefined (black)

Clone the repo

git clone https://github.com/pagraf/MagicBathyNet.git

Installation Guide

The requirements are easily installed via Anaconda (recommended):

conda env create -f environment.yml

After the installation is completed, activate the environment:

conda activate magicbathynet

Open Jupyter Notebook:

jupyter notebook

Train and Test the models

To train and test the bathymetry models use MagicBathyNet_bathymetry.ipynb.

To train and test the pixel-based classification models use MagicBathyNet_pixelclass.ipynb.

Pre-trained Deep Learning Models

We provide code and model weights for the following deep learning models that have been pre-trained on MagicBathyNet for pixel-based classification and bathymetry tasks:

Pixel-based classification

Learning-based Bathymetry

To achieve the results presented in the paper, use the parameters and the specific train-evaluation splits provided in the dataset. Parameters can be found here while train-evaluation splits are included in the dataset.

Example testing results

Example patch of the Agia Napa area (left), pixel classification results obtained by U-Net (middle) and predicted bathymetry obtained by MagicBathy-U-Net (right). For more information on the results and accuracy achieved read our paper.

Authors

Panagiotis Agrafiotis http://users.ntua.gr/pagraf/

Feedback

Feel free to give feedback, by sending an email to: pagraf@central.ntua.gr

Funding

This work is part of MagicBathy project funded by the European Union’s HORIZON Europe research and innovation programme under the Marie Skłodowska-Curie GA 101063294. Work has been carried out at the Remote Sensing Image Analysis group. For more information about the project visit https://www.magicbathy.eu/.