Datasets for Architectural Research: A review of selected plan-based datasets and their implementations

Architectural research increasingly relies on plan-based datasets to support various applications of computational design and machine learning tools, such as structural recognition, plan generation, and 3D scene reconstruction. However, there is a lack of understanding and comparison of existing datasets' characteristics, potentials, and challenges. This thesis aims to address this gap by conducting a selective review of existing plan-based datasets and their implementations for architectural research. The thesis examines the data collection, annotation, analysis, and limitations of selected datasets with their usage and evaluation. The ultimate goal is to suggest best practices and future directions for architectural research using plan-based datasets. This thesis aims to contribute to the current research by providing a better understanding of the potential of these datasets to support architectural research. By providing potential for data format organization and proposing a collaborative data annotation process from the findings of the dataset review, the work can provide insights for researchers and practitioners interested in exploring the possibilities of plan-based datasets for computational design.