These rivers are not real by Muhammed Sit

Sit, Muhammed

Graduate Student

Civil and Environmental Engineering
IIHR--Hydroscience & Engineering

Hydroinformatics

2020-03-04

2020 submissions

This image is a set of outputs from a project where we investigate an application of image generation for river satellite imagery. In that project, we, specifically, propose a generative adversarial network (GAN) model capable of generating high-resolution and realistic river images that can be used to support models in surface water estimation, river meandering, wetland loss and other hydrological research studies. First, we summarized an augmented, diverse repository of overhead river images to be used in training. Second, we incorporate the Progressive Growing GAN (PGGAN), a network architecture that iteratively trains smaller-resolution GANs to gradually build up to a very high resolution, to generate 256x256 river satellite imagery. With conventional GAN architectures, difficulties soon arise in terms of exponential increase of training time and vanishing/exploding gradient issues, which the PGGAN implementation seems to significantly reduce. All in all we are able to generate satellite imagery where real looking rivers, although they are fake, are visible. This image is a collage of some images we generated.

Honorable Mention in the 2020 People's Choice Category

satellite
imagery
river

Capture Your Research

University of Iowa. Lichtenberger Engineering Library
University of Iowa. College of Engineering. NEXUS Program
Virgil M. Hancher Auditorium

Attribution-NonCommercial 2.0 (CC BY-NC 2.0) https://creativecommons.org/licenses/by-nc/2.0/

Contact Kari Kozak in the Litchenberger Engineering Library at the University of Iowa: http://www.lib.uiowa.edu/eng/contact/

These rivers are not real (Vertical).jpg

Still Image

scientific illustrations (images)