A Deep Learning Approach to Analyzing and Evaluating Historic Urban Landscapes

Abstract

Abstract The integration of new buildings into historic urban landscapes (HULs) is a complex task that involves input from multiple stakeholders with varying perspectives. This often leads to a time-consuming and contentious decision-making process due to the lack of reliable data and the reliance on personal interpretations. Previous studies have used deep learning methods in combination with expert evaluations to assess the quality of building integration in HULs. However, these studies have been found to be biased and have not succeeded in simplifying the decision-making process. The current study aims to introduce a new approach to the analysis of different urban spaces that addresses the limitations of traditional methods by using minimal human intervention. The study utilizes a comparative approach algorithm that compares two different HUL areas, and the resulting data are analyzed using a heatmap visualization method. The study concludes that this approach produces more accurate and unbiased results than traditional methods, and significantly reduces the time required for analysis. By identifying key elements of the HUL that define the area's identity, this method provides valuable information for urban planners and architects to make informed decisions when working within HULs.