This repository contains code related to the study analyzed in the paper cited below:
Kristollari, Viktoria, and Vassilia Karathanassi. "Convolutional neural networks for detecting challenging cases in cloud masking using Sentinel-2 imagery." In Eighth International Conference on Remote Sensing and Geoinformation of the Environment (RSCy2020), vol. 11524, p. 115240K. International Society for Optics and Photonics, 2020.
It can be accessed in: https://doi.org/10.1117/12.2571111
Run:
the commands in the "gdal_commands" file to convert the Sentinel-2 images to .tiff files and merge bands.
"data_downloading.py" to download the Sentinel-2 images.
"data_preprocessing.py" to create training and test data for the CNN.
"generator.py" to generate training and test patches.
"CNN_training.py" to train the CNN and save the weights.
"CNN_predictions.py" to create and save the predicted cloud masks.
"evaluation_metrics.py" to evaluate the cloud masks.
Detailed guidelines are included inside each script.
Run:
- "add_padding.py" to add zero padding to the Sentinel-2 images.
If you use this code, please cite the below paper.
@inproceedings{10.1117/12.2571111,
author = {Viktoria Kristollari and Vassilia Karathanassi},
title = {{Convolutional neural networks for detecting challenging cases in cloud masking using Sentinel-2 imagery}},
volume = {11524},
booktitle = {Eighth International Conference on Remote Sensing and Geoinformation of the Environment (RSCy2020)},
editor = {Kyriacos Themistocleous and Giorgos Papadavid and Silas Michaelides and Vincent Ambrosia and Diofantos G. Hadjimitsis},
organization = {International Society for Optics and Photonics},
publisher = {SPIE},
pages = {188 -- 201},
keywords = {Convolutional neural networks, Cloud masking, Sentinel satellite imagery, Thin cloud detection, Bright surfaces detection, Feature maps},
year = {2020},
doi = {10.1117/12.2571111},
URL = {https://doi.org/10.1117/12.2571111}
}