Road Extraction using Deep Learning

Abstract

The Road extraction from aerial image, stands as a quintessential node for the development of rudimentary layers in innumerable fields. From GIS, to Unmanned Aerial vehicles, road maps pave the foundation for data accumulation. This significant process is a result of number of mechanisms devised over the years through iterative experiments and research. However, the glut of methods available often pose as a hurdle in the selection process. In this project we implement a novel approach to solve the extraction problem, by incorporating generative algorithm using conditional adversarial networks. We investigate conditional adversarial networks as a general-purpose solution to image-to-image translation problems. These networks not only learn the mapping from input image to output image, but also learn a loss function to train this mapping. This makes it possible to apply the same generic approach to problems that traditionally would require very different loss formulations. The U-Network incorporated essentially convolves and de-convolves over the generative model, thus producing a pixel to pixel image translation, the result of which is the vector road map of its corresponding aerial image. The entire model is trained on a 990 MS GPU for computational ease.