Measuring the Degree of Beauty based on Alexander's 15 Properties of Living Structure
This supplementary dataset accompanies the Master's thesis, Characterizing Architectural and Urban Beauty from the Perspective of Living Structure Using AI and Geospatial Big Data, authored by Andy Jingqian Xue at HKUST(GZ). The dataset is organized into three distinct parts:
1. ScenicOrNot Dataset: This collection of urban and architectural images is designed to validate the novel Large Language Model (LLM) approach for quantifying aesthetic beauty, as proposed in the thesis.
- architectual_images.rar (1.4GB): Subset focusing on architectural images
- all_images.csv (16MB): Comprehensive ratings and annotations for all images
- archi_images.csv (1.4MB): Specific ratings and annotations for architectural images
- 1000_pairs.csv (15KB): Contains 1000 image pairs for comparative analysis
Note: Due to their large size, all images can be downloaded using the URLs provided in all_images.csv.
2. Architectural Style Dataset: This dataset facilitates a comparative study of architectural beauty, specifically examining the perceived decline in aesthetic value from traditional Western architectural styles to modern and contemporary designs.
- images.rar (1.1GB): Collection of architectural images across different styles
- style_pairs.csv (275KB): IDs for architectural style comparisons, 1500 paris in total
3. Urban Function Dataset: This part provides data to analyze the heterogeneity of beauty across various urban forms and functional zones.
- images.zip (74MB): Collection of images representing different urban functions
The dataset serves as a comprehensive resource for studying urban and architectural beauty through the lens of living structure theory, combining traditional aesthetic analysis with modern AI approaches.
