Progressively growing generative adversarial networks for high resolution semantic segmentation of satellite images

Authors

Edward Collier, Louisiana State University
Kate Duffy, Northeastern University
Sangram Ganguly, NASA Ames Research Center
Geri Madanguit, BAER Institute
Subodh Kalia, BAER Institute
Gayaka Shreekant, BAER Institute
Ramakrishna Nemani, NASA Ames Research Center
Andrew Michaelis, NASA Ames Research Center
Shaung Li, NASA Ames Research Center
Auroop Ganguly, Northeastern University
Supratik Mukhopadhyay, Louisiana State University

Document Type

Conference Proceeding

Publication Date

7-2-2018

Abstract

Machine learning has proven to be useful in classification and segmentation of images. In this paper, we evaluate a training methodology for pixel-wise segmentation on high resolution satellite images using progressive growing of generative adversarial networks. We apply our model to segmenting building rooftops and compare these results to conventional methods for rooftop segmentation. We present our findings using the SpaceNet version 2dataset. Progressive GAN training achieved a test accuracy of 93% compared to 89% for traditional GAN training.

Publication Source (Journal or Book title)

IEEE International Conference on Data Mining Workshops Icdmw

First Page

763

Last Page

769

Recommended Citation

Collier, E., Duffy, K., Ganguly, S., Madanguit, G., Kalia, S., Shreekant, G., Nemani, R., Michaelis, A., Li, S., Ganguly, A., & Mukhopadhyay, S. (2018). Progressively growing generative adversarial networks for high resolution semantic segmentation of satellite images. IEEE International Conference on Data Mining Workshops Icdmw, 2018-November, 763-769. https://doi.org/10.1109/ICDMW.2018.00115