A geography of indoors for analyzing global ways of living using computer vision
摘要
Globalization is claimed to have a homogenizing effect, reducing pronounced local cultural differences. Indoor living spaces are among the most vivid expressions of local culture, yet they remain underexplored in this context. Our visual AI framework, utilizing a unique dataset of over 400,000 Airbnb images, investigates the diversity in living spaces across 80 cities. By employing deep learning classification models, Gradient-Weighted Class Activation Mapping techniques, and statistical analysis, we demonstrate that both geographic proximity and the extent of globalization significantly correlate with the visual characteristics of indoor spaces (R = 0.23–0.30 and R ≈ 0.47, respectively). Our results indicate that despite global pressures and trends towards cultural homogenization, local identities and cultural distinctions nevertheless remain.