SAR Image Gray Scale and Optical Dataset

The data includes a combination of SAR and optical (RGB) imagery that consists of Sentinel\u20111 and Sentinel\u20112 satellites, which were initially curated and made available by Technical University of Munich (TUM) with the help of Dr. Michael Schmitt. In the large data set, we have chosen the pairs of images recorded in autumn season representing four types of land cover, namely: barren land, grassland, agricultural land, and urban areas. The chosen pictures are very diverse within the classes in order to have diversity and abundance of representation.Such pairs of images are specifically good to train Conditional Generative Adversarial Networks (cGANs). Considering that the data have a complex nature and not even geometric regularity, this dataset is also a strong benchmark to test the generalization ability and robustness of deep learning models in a wide range of tasks.