Abstract

In recent years, machine learning has progressed in image and semantic recognition, providing digital platforms for architectural and landscape research. The cultural heritage sites along the Jiangnan Canal show great aesthetic value in their forms, which deserve research and preservation with digital technologies. This research collects images of cultural architectural heritage along Jiangnan Canal to create a dataset, including gardens, gates, pavilions, pagodas, residences, temples, bridges, and street blocks. Two deep learning models were developed on Google Teachable Machine for building type recognition and classification, and comparative experiments with historical and modern images were conducted to analyze the model’s accuracy. The success rate results show that thirty-nine out of sixty-four images (60.94%) passed the 77 percent accuracy test on Model A, while forty-eight out of sixty-four images (75%) passed the test on Model B. Classification results show the factors that affect the accuracy of the model and further adjustments of the model are proposed. This article contributes to the application of Artificial Intelligence in cultural heritage protection and offers new methods for recognition of historical architecture types. Using digital techniques and platforms, this classifier has the potential to be applied in a wide range of digital design areas and could be further developed for recognizing architectural morphology.