Lei and colleagues transform restorative perception from a small-sample psychological measure into a spatially extensive urban design variable. Using Shanghai’s Huangpu District, they integrate street-view imagery, restorative component scales, random forests, multiscale geographically weighted regression and space syntax to identify how streetscape elements influence perceived restoration. The iconic contribution is spatial heterogeneity: trees, sidewalks, buildings, walls and network accessibility do not exert uniform effects, and high restorative perception clusters around parks, waterfronts, strong infrastructure and well-designed buildings. The theoretical value lies in treating perception as geographically conditioned rather than as a universal reaction to visual features. Methodologically, the paper combines computational vision with spatial statistics and configurational analysis, allowing subjective environmental qualities to be mapped at urban scale. It bridges environmental psychology, urban analytics and design research by showing how machine-assisted measurement can reveal location-specific perceptual patterns while retaining the importance of context, accessibility and the material composition of streets.