A Structured Repository of Architectural Visualisation Metadata, Visual Characteristics, and Machine-Vision Descriptions

This dataset consists of a comprehensive structured spreadsheet containing 296 entries that document the intersection of architectural visualisation, sustainability claims, and artificial intelligence interpretation. The data was compiled through a rigorous content analysis of secondary sources from leading digital architectural platforms such as ArchDaily and Architizer. It serves as a multidimensional record that bridges the gap between technical visual metadata and qualitative design narratives. The repository is organised into several key thematic clusters. The first section contains essential project metadata, including project titles, architectural firms, geographic locations, and scales ranging from individual building components to urban districts. This allows researchers to categorise the data based on global trends and architectural typologies. The second cluster focuses on the characteristics of the visual media itself. It records static image, the specific visualisation type such as 3D renders or photography, and the artistic point of view. Furthermore, it incorporates detailed perceptual variables including brightness, tonal range, colour temperature, saturation, and composition. These metrics provide a quantitative basis for studying the atmospheric and technical qualities of architectural imagery. A significant portion of the dataset is dedicated to sustainability. It captures the specific claims made by architects and developers regarding environmental performance. This is further broken down into sustainable design attributes, such as passive cooling or renewable energy systems, and the specific sustainable materials highlighted in the visuals. By categorising these elements, the data enables an investigation into how environmental responsibility is communicated visually to the public. A unique feature of this dataset is the integration of AI-based visual inspection results. Each entry includes descriptive tags and adjectives derived from AI image-to-text processes. This allows for a direct comparison between human-authored project descriptions and the objective interpretations provided by computer vision models. The dataset is provided as a single Excel file to maintain the integrity of the cross-referenced variables. Manual untuk menggunakan dataset ini tertulis dalam A Structured Repository of Architectural Visualisation Annotation Manual (PDF). Users should note that while the metadata is exhaustive, the images themselves are referenced via permalinks to respect original copyright holders. Meanwhile, high resolution charts for some key features are in the Distribution Charts folder. This collection represents a significant contribution to the digitisation of architectural research and the study of human-centric sustainable environments.