Remote Sensing Image Super-Resolution via Saliency-Guided Feedback GANs

January 6, 2021 ยท View on GitHub

by Hanlin Wu, Libao Zhang, and Jie Ma, details are in paper.

Introduction

This repository is build for the proposed SG-FBGAN, which contains full training and testing code.

framework

Usage

Clone the repository:

git clone https://github.com/BNUAI/SG-FBGAN.git

Requirement:

  • tensorflow==1.14.0
  • tensorlayer==1.11.0
  • numpy
  • easydict
  • opencv-python
  • tqdm
  • wget
pip install -r requirements.txt

Test with our pre-trained models:

  1. Download the pre-trained SG-FBGAN models.
  1. Unzip the the downloaded file, and put the pre-trained model on path : experiments/exp_name
  2. Do testing: 
    python predict.py --opt exp_name
    Note: The GeoEye-1 dataset will be downloaded automatically. If the download fails, please download it manually from here, and then put the downloaded file on path : data/sr_geo.npz.

Train with our GeoEye dataset:

python train.py --opt config/va_fbgan_x3_BI.json

Train with your own dataset:

  1. change the datapath and savepath in data_loader/make_npz.py, and then make the .npz file:

    python data_loader/make_npz.py

  2. change the data_path in config/your_own_config_file.json.

  3. Do training:

    python train.py --opt config/your_own_config_file.json

Cite

H. Wu, L. Zhang and J. Ma, "Remote Sensing Image Super-Resolution via Saliency-Guided Feedback GANs," in IEEE Transactions on Geoscience and Remote Sensing, doi: 10.1109/TGRS.2020.3042515.

Contact