Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
July 27, 2026Computers Environment and Urban SystemsOpen Access

Deep learning reveals cross-platform consistency in street view imagery for urban perception mapping

View Full Paper
Ask AI
Bookmark
Share

Authors

TBTingting BaiDXDong XuFGFeng Gao

Discussion

Loading...

Member takes

Overview

Randomized trial tests urban perception mapping with street view imagery from Baidu and Google, suggesting effective data integration methods.

Key Points

  • This research aims to evaluate the consistency of urban perception mapping using street view imagery from different platforms.
  • Evaluated paired images from Baidu Street View and Google Street View in Hong Kong.
  • Employed stacking ensemble learning, SHAP-based interpretation, and Mantel correlation analysis at both pixel and sampling-point scales.
  • Conducted robustness tests with various segmentation pipelines and perception-backbone models.
  • At the sampling-point scale, correlations for vegetation, buildings, and sky were 0.753, 0.756, and 0.720 respectively.
  • At the pixel scale, vegetation distributions showed a mean correlation of 0.84, significant at p < 0.001; ‘Beauty’ correlated positively at 0.645.
  • Natural elements positively influence perceived beauty, while artificial features link to negative perceptions.

Cite This Study

Bai et al. (2026) studied this question.

synapsesocial.com/papers/6a6700af40bca442e0d4a8dahttps://doi.org/10.1016/j.compenvurbsys.2026.102493
View Full Paper
Ask AI
Bookmark
Share