Kunye-Shen / EMFINet

Code for TGRS 2021 paper. Edge-Aware Multiscale Feature Integration Network for Salient Object Detection in Optical Remote Sensing Images.

EMFINet (TGRS 2021)

'Edge-Aware Multiscale Feature Integration Network for Salient Object Detection in Optical Remote Sensing Images', Xiaofei Zhou, Kunye Shen, Zhi Liu, Chen Gong, Jiyong Zhang, and Chenggang Yan.

Required libraries

Python 3.7
numpy 1.18.1
scikit-image 0.16.2
PyTorch 1.4.0
torchvision 0.5.0
glob

The SSIM loss is adapted from pytorch-ssim.

Usage

1. Clone this repo

https://github.com/Kunye-Shen/EMFINet.git

2. We also provide the predicted saliency maps (GoogleDrive or baidu extraction code: 9nje.). You can download directly through the above methods, or contact us through the following email.

[e-mail]

Architecture

Quantitative Comparison

Qualitative Comparison

ORSSD

EORSSD

Citation

@article{zhou2021edge,
  title={Edge-Aware Multiscale Feature Integration Network for Salient Object Detection in Optical Remote Sensing Images},
  author={Zhou, Xiaofei and Shen, Kunye and Liu, Zhi and Gong, Chen and Zhang, Jiyong and Yan, Chenggang},
  journal={IEEE Transactions on Geoscience and Remote Sensing},
  year={2021},
  publisher={IEEE}
}