MDFN for Short-Term Multitemporal HR Images Classification

It is the python-keras implementation of the paper: A Novel Multitemporal Deep Fusion Network (MDFN) for Short-Term Multitemporal HR Images Classification

For more information, please see our published paper at IEEE JSTARS

Prerequisites

Windows 10
Python 3.6
CPU or NVIDIA GPU
CUDA 9.0
Keras 2.2.4

Quick Start

You can run a demo to get started.

python MDFN_demo.py

Prepare Datasets

Using other dataset mode

In this case, the data structure should be the following:

"""
Image classification data set with pixel-level binary labels；
├─Image_T1 & Image_T2 & Image_T3...
├─label
├─train_set
└─test_set
"""

Citation

If you use this code for your research, please cite our papers.

@ARTICLE{9573468,  
author={Zheng, Yongjie and Liu, Sicong and Du, Qian and Zhao, Hui and Tong, Xiaohua and Dalponte, Michele},  
journal={IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing},   
title={A Novel Multitemporal Deep Fusion Network (MDFN) for Short-Term Multitemporal HR Images Classification},   
year={2021},  
volume={14},  
number={},  
pages={10691-10704},  
doi={10.1109/JSTARS.2021.3119942}
}

Acknowledgments

Our code is inspired by keras-SSUN.