Integrated Deep learning framework for visual perception assessment of historic dwelling exteriors

Exploring perceptual evaluations of historic dwellings is essential for shaping the built environment and conserving cultural heritage, yet traditional evaluation methods often lack sufficient depth and adaptability. This study integrates questionnaire surveys with deep learning methods to examine the associations between exterior features of historic dwellings and perceptual evaluations from an expert perspective, and to assess facade perception in real traditional settlements. The results reveal consistent associations among complexity, perceived value, order, pleasure, and comfort, while boredom shows an opposite tendency. Window-related attributes, color complexity, and structural types exhibit stable directional patterns across perceptual dimensions. Perceptual differences among dwellings suggest remaining potential for optimization in conservation and renovation.