FACA (Fine-grained Attributes for Cultural Architecture)

FACA-1K (Fine-grained Attributes for Cultural Architecture) is an image\u2013knowledge paired dataset constructed to advance fine-grained, culturally aware captioning. The dataset comprises 1,000 high-resolution photographs collected from 204 traditional Korean heritage buildings officially designated by the Cultural Heritage Administration, each image paired with structured architectural attributes derived by experts. Knowledge entries are compiled from authoritative sources, such as national heritage archives and encyclopedic records, translated into English, and organized into a domain schema that encompasses roof types, column styles, bracket systems, plan\/form, decorative elements, and country cues. The images capture both overall views and detailed components, ensuring that visual evidence aligns with architectural terminology.