Application of a Deep Learning Approach to Map and Predict the Age-friendliness of the Built Environment

Publication Type:
Issue Date:
Full metadata record
Population aging is one of the most significant phenomena worldwide. The concept of age-friendly cities is gaining attention globally, considering the increase in the aging rate. Research into age-inclusive built environments is a relatively evolving area. However, there is a lack of extensive research on urban data acquisition on cities' age-friendly features using modern technologies. Considering the size of modern cities, it is beyond the capacity of traditional audit tools to develop a comprehensive analysis concerning spatial issues. Therefore, this research's core objective is to predict the built environment's age-friendliness using modern computer vision tools to overcome traditional auditing's limitations. Urban computing and deep neural networks are employed to map, analyze, and predict the urban environment's age-friendliness utilizing street-level images. The model is built using a transfer learning technique based on a pre-trained architecture (VGG-16), enabling a more rapid and precise analysis of the Google Street View images from three Sydney neighbourhoods. The proposed model is scalable to more neighbourhoods, and urban planners can apply it to evaluate neighbourhood conditions. The proposed model can achieve adequate accuracy in generating human-like assessments. Ultimately, some mitigation measures for developing age-friendly urban places will be offered based on the study's findings.
Please use this identifier to cite or link to this item: