Skip to content
/ RDPNet Public

RDP-Net: Region Detail Preserving Network for Change Detection

License

Notifications You must be signed in to change notification settings

Chnja/RDPNet

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

14 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

RDP-Net: Region Detail Preserving Network for Change Detection

author Frame license Paper

PWC

RDP-Net

The pytorch implementation for RDP-Net.

from RDPNet import RDPNet
net = RDPNet(in_ch=3, out_ch=2).to(device)

The pytorch implementation for Edge Loss.

from edgeloss import combine_loss
loss = combine_loss(prediction, target)

Well-trained model.

# For CDD dataset
net.load_state_dict(torch.load("RDPNet_CDD.pth"))

Paper

This work has been published on IEEE Transactions on Geoscience and Remote Sensing (TGRS).

The paper is now avaliable on TGRS and arXiv Link.

@article{chen2022rdpnet,
    author={Chen, Hongjia and Pu, Fangling and Yang, Rui and Tang, Rui and Xu, Xin},
    journal={IEEE Transactions on Geoscience and Remote Sensing}, 
    title={RDP-Net: Region Detail Preserving Network for Change Detection}, 
    year={2022},
    volume={60},
    number={},
    pages={1-10},
    doi={10.1109/TGRS.2022.3227098},
}

Introduction

Change detection (CD) is an essential earth observation technique. It captures the dynamic information of land objects. With the rise of deep learning, neural networks (NN) have shown great potential in CD. However, current NN models introduce backbone architectures that lose detailed information during learning. Moreover, current NN models are heavy in parameters, which prevents their deployment on edge devices such as UAVs. In this work, we tackle this issue by proposing RDP-Net: a region detail preserving network for CD. We propose an efficient training strategy that constructs the training tasks during the warmup period of NN training and lets the NN learn from easy to hard. The training strategy enables NN to learn more powerful features with fewer FLOPs and achieve better performance. Next, we propose an effective edge loss that increases the penalty for errors on details and improves the network's attention to details such as boundary regions and small areas. Furthermore, we provide a NN model with a brand new backbone that achieves the state-of-the-art empirical performance in CD with only 1.70M parameters. We hope our RDP-Net would benefit the practical CD applications on compact devices and could inspire more people to bring change detection to a new level with the efficient training strategy.

Dataset

CDD

Paper: Change Detection in Remote Sensing Images using Conditional Adversarial Networks

Link: is not avaliable now, but you can contact the author ([email protected]) to get this Dataset.

LEVIR-CD

Paper: A Spatial-Temporal Attention-Based Method and a New Dataset for Remote Sensing Image Change Detection

Link

Comparison Methods

Contact

Hongjia Chen: [email protected]

About

RDP-Net: Region Detail Preserving Network for Change Detection

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages