Skip to main navigation Skip to search Skip to main content

Daytime to Nighttime Street View Image Generation for 24-h Safety Perception Mapping

  • Jiajing Dai
  • , Zhiyi Liu
  • , Tingting Li
  • , Tianyi Ren
  • , Waishan Qiu
  • , Da Chen
  • , Wenjing Li
  • University of Pennsylvania Stuart Weitzman School of Design
  • Beijing University of Civil Engineering and Architecture
  • South Minzu University
  • Smart Gwei Tech
  • Chinese University of Hong Kong
  • University of Tokyo

Research output: Chapter or section in a book/report/conference proceedingBook chapter

1   Link opens in a new tab Citation (SciVal)

Abstract

Urban safety perception is crucial for quality of life, yet research on nighttime safety perception remains limited due to the lack of nighttime street view images (SVIs). This study uses generative models to translate daytime SVIs into nighttime ones (D2N), addressing this data gap. However, challenges exist in understanding how micro-level streetscape features affect nighttime perceived safety. To fill these gaps, this study collected 1,461 paired SVIs from high-density areas and 222 from low-density areas across four Chinese cities and Boston. We validated the high-density D2N (HD2N) and low-density D2N (LD2N) models using standard metrics and human-machine adversarial scoring. Our findings show that: (1) Urban density affects D2N translation accuracy, with higher densities posing challenges; (2) HD2N requires over 760 samples to converge; (3) Roads improve safety perception, while trees reduce it. These results offer insights for urban planning and safety improvements.

Original languageEnglish
Title of host publicationDigital-Era Urban Transformations
Subtitle of host publicationAdvancements in Data Science, Analytics and Technology. CUPUM 2025
EditorsR. Goodspeed, E. Suel, H. Chen, J. Barros, C. Pettit
Place of PublicationCham, Switzerland
PublisherSpringer
Chapter15
Pages263-281
Number of pages19
Edition1st
ISBN (Electronic)9783031983009
ISBN (Print)9783031982996
DOIs
Publication statusPublished - 16 Dec 2025

Publication series

NameUrban Book Series
VolumePart F1191
ISSN (Print)2365-757X
ISSN (Electronic)2365-7588

Keywords

  • Day and night discrepancy
  • Day-to-night translation
  • Nighttime safety perception
  • Street view image
  • Urban environment auditing

ASJC Scopus subject areas

  • Geography, Planning and Development
  • Urban Studies

Fingerprint

Dive into the research topics of 'Daytime to Nighttime Street View Image Generation for 24-h Safety Perception Mapping'. Together they form a unique fingerprint.

Cite this