This repo is the official implementation of Weakly supervised high spatial resolution land cover mapping based on self-training with weighted pseudo-labels. The code will be release public in June 2023.
The models with the scores can be downloaded from Baidu Cloud (Extraction code: 1234).
conda install pytorch==1.11.0 torchvision==0.12.0 torchaudio==0.11.0 cudatoolkit=11.3 -c pytorch
pip install mmcv-full==1.5.1 -f https://download.openmmlab.com/mmcv/dist/cu113/torch1.11.0/index.html
pip install mmsegmentation==0.24.1
pip install ever-beta==0.2.3
pip install timm
pip install --upgrade git+https://github.com/Z-Zheng/SimpleCV.gitln -s </path/to/LoveDA> ./LoveDA- Stage one, training with source domain:
python source.py - Stage two, self-training:
python train.py - the oracle setting to test the upper limit of our method’s accuracy in a single domain:
python oracle.py
This code is heavily borrowed from LoveDA.
If you find this repo useful for your research, please consider citing the paper as follows:
@article{liu2022weakly,
title={Weakly supervised high spatial resolution land cover mapping based on self-training with weighted pseudo-labels},
author={Liu, Wei and Liu, Jiawei and Luo, Zhipeng and Zhang, Hongbin and Gao, Kyle and Li, Jonathan},
journal={Int. J. Appl. Earth Obs. Geoinf.},
volume={112},
pages={102931},
year={2022}
}