Authors: Yifan Sun*, University of Washington, Bo Zhao, University of Washington
Topics: Geographic Information Science and Systems, Urban Geography, Cartography
Keywords: geographical imagination, sense of place, street view image, semantic segmentation, generative adversarial network (GAN)
Session Type: Virtual Paper
Start / End Time: 8:00 AM / 9:15 AM
Room: Virtual 32
Presentation File: No File Uploaded
When people wander in the space and time of daily life, the imagination of unknown places never stops. The exploration of exotic customs gives people aesthetic enjoyment, the vision of prosperous cities makes people full of fighting, and the concern about war and poverty makes people full of sympathy and reflection on society. Can the overlap of geographical imagination and everyday places make people understand places more deeply? To demonstrate the question, this paper proposes a framework for recreating different local scenes in familiar places, and explores the potential of imaginary places in aesthetic and social reflection through three examples. To be specific, this paper firstly realizes the transformation from Google street view image to semantic layout image through semantic segmentation technology, and then captures the predominant visual properties of street view images in different places through GANs, and further reconstruct the visual features of different places in a specific semantic layout image. The three examples of this paper are the oriental Seattle, the brand-new Mumbai slum, and the war-torn Wuhan. The empirical results show that the framework proposed in this paper can reproduce the visual images of different location in specific places in a satisfactory way to promote people's understanding of places. Through this paper, we not only hope to expand the connotation of geographical imagination derived from human geography, but also hope to bring enlightenment to GIScientists in their discussion of geographic imagination system.