Thesis

Unsupervised Super Resolution for Sentinel-2 satellite imagery

In this work three unsupervised deep learning models were utilized for Super Resolving satellite imagery obtained from Sentinel-2 constellation.

1. Deep Image Prior (DIP)

The original implementation of this model was proposed by Ulyanov et al.

Requirments

• python == 3.8
• earthpy
• numpy
• pytorch
• matplotlib
• scikit-image
• gdal
• rasterio
• jupyter notebook

2. Zero-Shot Super Resolution (ΖSSR)

The original implementation of this model was proposed by Shocher et al.

Requirments

• python == 2.7
• numpy
• tensorflow
• matplotlib
• scikit-image
• opencv-python
• imageio

3. Degradation-Aware Super Resolution (DASR)

The original implementation of this modes was proposed by Wang et al.

Requirments

• Python 3.6
• PyTorch == 1.1.0
• numpy
• skimage
• imageio
• matplotlib
• cv2

Train

1. Prepare training data

1.1 Download the DIV2K dataset and the Flickr2K dataset.

1.2 Combine the HR images from these two datasets in your_data_path/DF2K/HR to build the DF2K dataset.

2. Begin to train

Run ./main.sh to train on the DF2K dataset. Please update dir_data in the bash file as your_data_path.

Test

1. Prepare test data

Download benchmark datasets (e.g., Set5, Set14 and other test sets) and prepare HR/LR images in your_data_path/benchmark.

2. Begin to test

Run ./test.sh to test on benchmark datasets. Please update dir_data in the bash file as your_data_path.

Quick Test on An LR Image

Run ./quick_test.sh to test on an LR image. Please update img_dir in the bash file as your_img_path.