Abstract
Social inclusion has become an important issue for cultural heritage planning in the past decade. Whereas the Recommendation of Historic Urban Landscape (HUL) called for tools for civic engagement and knowledge documents, social media already function as a platform for online communities to actively involve themselves in heritage-related activities. The aim of this research is to understand and explain the heritage attributes revealed in Flickr social media posts in a worldheritage city using artificial intelligence (AI) tools, as continuation to a previous study. Using IBM Watson Visual Recognition API, a computer vision model on classifying heritage attributes is trained with a total of 914 images. The model is evaluated on the test set with machine learning metrics such as precision, recall, and F1. A macro-average F1 of 0.76 is reached using the model, proving that it is feasible and reliable for further applications. Error analysis implies that the applied category of heritage attribute needs to be further clarified, since some classes are not mutually exclusive, making the model hard to perform well in those specific classes. This research shows that AI tools are helpful to automate the heritage attribute identification process on social media posts. Such tools can potentially help heritage researchers and practitioners to understand the opinions of general public, which can stimulate social inclusion in the heritage planning and management.
Presenters
Nan BaiPhD Candidate, Architecture Engineering and Technology, TU Delft, Zuid-Holland, Netherlands Ana Roders
Pirouz Nourian
Details
Presentation Type
Theme
KEYWORDS
Social Media, Social Inclusion, Computer Vision, Artificial Intelligence, Historic Urban