Historic townscapes represent invaluable cultural heritage, with building facades serving as primary carriers of architectural character. However, traditional documentation relies on resource-intensive manual surveys, while existing digital archives often lack systematic organization, element-level analysis, and chronological tracking capabilities. This research proposed a framework for an automated facade documentation that converts Street View Image into structured, analyzable architectural elements data. First, automated scripts acquire geo-tagged Street View Images within conservation zones. Then, a Convolutional Neural Network segments individual façades and classify architectural elements such as roofs, windows, doors, and shopfronts. Finaly, extracted element instances, along with their geometric outlines and attributes, are integrated into a relational database for longitudinal analysis. Training and validation losses declined steadily while precision, recall, and mAP metrics increased across 1000 epochs, reaching stable performance (~0.55–0.70). These results confirm that the model learned façade features effectively and is suitable for large-scale documentation. Implementation in case study towns demonstrates advantages over traditional methods: the ability to quantitatively analyze partial architectural patterns over the years and efficiency of image processing with minimal human intervention. Ultimately, this research contributes to a structured facade typology system, and an open-source implementation that transforms passive image collections into tangible, analyzable data. The resulting platform enables urban planners, researchers, and community stakeholders with quantitative data on architectural evolution, supporting evidence-based preservation effort and deeper understanding townscape heritage.