Retail Façade Change Detection via Street-level Imagery

Retail stores are a primary setting for urban consumption, playing a crucial role in shaping urban economic vitality. However, current research on the quantitative study of retail façade environments and their spatio-temporal evolution remains limited. This study proposes a retail façade representation framework based on street level imagery, combining viewpoint alignment with three complementary models—DINOv3, SegFormer, and CIELAB—to capture the multidimensional semantic and visual features of retail façades. Using Liverpool, one of the UK’s leading and economically vibrant cities, as an example, the framework identifies spatio-temporal patterns of retail transformation and renewal hotspots, and systematically compares the performance of the three modelling approaches. The findings provide valuable insights into urban retail dynamics and offer methodological support for urban and business analysis.