4x and 10x Super Resolution Generator Models Trained With Planet CubeSat Satellite Imagery

Final resampling generator models produced from the Enhanced Super Resolution Generative Adversarial Network (ESRGAN) (https://github.com/xinntao/ESRGAN). ESRGAN was trained at two different resampling factors, 4x and 10x, using a training data set of global Planet CubeSat satellite images. These generators can be used to resample Planet CubeSat satellite images from 30m and 12m to 3m resolution. Descriptions and results of training can be found at https://wandb.ai/elezine/pixelsmasher. In press at Canadian Journal of Remote Sensing: Super-resolution surface water mapping on the Canadian Shield using Planet CubeSat images and a Generative Adversarial Network, Ekaterina M. D. Lezine, Ethan D. Kyzivat, and Laurence C. Smith (2021).