This repository contains the codes for the paper "CloudSeg: A Multi-Modal Learning Framework for Robust Land Cover Mapping under Cloudy Conditions"
If you use the codes for your research, please cite us accordingly:
@article{xu2024cloudseg,
title={CloudSeg: A multi-modal learning framework for robust land cover mapping under cloudy conditions},
author={Xu, Fang and Shi, Yilei and Yang, Wen and Xia, Gui-Song and Zhu, Xiao Xiang},
journal={ISPRS Journal of Photogrammetry and Remote Sensing},
volume={214},
pages={21--32},
year={2024}
}
This code has been tested with CUDA 11.7 and Python 3.8.
conda create -n CloudSeg python=3.8
conda activate CloudSeg
pip install torch==2.0.0 torchvision==0.15.1 torchaudio==2.0.1
pip install rasterio tqdm timm six scikit-learn
pip install pretrainedmodels efficientnet_pytorch
You can download the pretrained model (TeacherNet.pth & StudentNet.pth) and put it in './checkpoints'.
Use the following command to test the network:
cd ./StudentNet
python test_SS.py
Use the following command to train the network:
'''
1. Train the Teacher network
'''
cd ./TeacherNet
python train_SS.py
'''
2. Train the Student network
'''
cd ./StudentNet
python train_SS.py
Our experiments are conducted on two benchmark datasets: M3M-CR and WHU-OPT-SAR. The M3M-CR dataset features cloud-covered optical images derived from real remote sensing scenarios, while the WHU-OPT-SAR dataset does not include cloud-covered images corresponding to its cloud-free counterparts. We perform artificial cloud layer synthesis on the available cloud-free images within the WHU-OPT-SAR dataset to simulate the effect of cloud cover.
We are glad to hear if you have any suggestions and questions.
Please send email to xufang@whu.edu.cn