mila-ai4h/mid-space

MID-Space数据集旨在将AI生成的城市公共空间可视化与蒙特利尔多样化和边缘化社区的偏好对齐。它包含文本提示、Stable Diffusion XL图像和基于六个标准的社区偏好注释：可访问性、安全性、多样性、包容性、邀请性和舒适性。数据集包括CSV文件、图像文件夹和文档文件，CSV文件包含35,510行数据，图像文件夹包含约13,465张JPG图像。数据收集过程包括标准选择、注释者选择、提示收集、图像生成和人工注释。

The MID-Space dataset is designed to align AI-generated visualizations of urban public spaces with the preferences of diverse and marginalized communities in Montreal. It includes textual prompts, Stable Diffusion XL images, and annotations reflecting community preferences based on six criteria: Accessibility, Safety, Diversity, Inclusivity, Invitingness, and Comfort. The dataset structure includes CSV files with preferences, image folders, and documentation files. The data collection process involved criteria selection, annotator selection, prompt collection, image generation, and human annotation. The dataset is intended for use in research and can be loaded using Python libraries. The README also provides installation instructions, citation information, acknowledgements, and contact details.