Annotated Image Database of Architecture-CNNs (AIDA-CNNs)

By leveraging the state-of-the-art deep-learning-based classification models, a series of hierarchical multi-label classification models (AIDA-CNNs) are provided as a baseline for the task of architectural category classification. With the high-diversity and high-coverage of the database, AIDA along with the AIDA-CNNs offers a novel resource to guide future works on the evaluation of the performance of existing machine learning models with respect to architectural design research, as well as encouraging new ones.