Supporting data for "CIM-WV: A 2D semantic segmentation dataset of rich window view contents in high-rise, high-density Hong Kong based on photorealistic City Information Models"

This is the official repository of the CIM-WV dataset. For technical details, please refer to:Li, M., Yeh, A. G. &amp; Xue, F. (2023). CIM-WV: A 2D semantic segmentation dataset of rich window view contents in high-rise, high-density Hong Kong based on photorealistic City Information Models. Urban Informatics, 1-24.This study was supported in part by the Department of Science and Technology of Guangdong Province (GDST) (2020B1212030009, 2023A1515010757) and the University of Hong Kong (203720465).<b>Overview of CIM-WV</b>This paper presents a City Information Model (CIM)-generated Window View (CIM-WV) dataset comprising 2,000 annotated images collected in the high-rise, high-density urban areas of Hong Kong. 1) Window view images of CIM-WV depict diversified urban scenes of Hong Kong at different locations, elevations, and orientations2) The CIM-WV includes seven semantic labels, i.e., building, sky, vegetation, road, waterbody, vehicle, and terrain.In addition, we provide variants of DeepLab V3+ models trained on CIM-WV, real window view images, Google Earth CIM-generated window view images from New York, and Google Earth CIM-generated window view images from Singapore, respectively.You can modify the source code here to use the trained DeepLab V3+ models. <b>Contribution</b>1) CIM-WV is the first public CIM-generated photorealistic window view dataset with rich semantics. 2) Comparative analysis shows a more accurate window view assessment using deep learning from CIM-WV than deep transfer learning from ground-level views.3) For urban researchers and practitioners, our publicly accessible deep learning models trained on CIM-WV enable novel multi-source window view-based urban applications including precise real estate valuation, improvement of built environment, and window view-related urban analytics.Please cite our paper and dataset, if you find our work useful for your research and practices. Many thanks.For any inquiries, please feel free to contact Maosu at [e-mail] or Dr. Frank at [e-mail].

本仓库为CIM-WV数据集的官方代码托管仓库。如需了解技术细节，请参阅：Li M、Yeh A G与Xue F于2023年发表的论文《CIM-WV：基于照片级真实感城市信息模型（City Information Model, CIM）构建的香港高密度高层城区富语义窗口景观二维语义分割数据集》，刊载于《城市信息学》（*Urban Informatics*），页码1-24。本研究得到广东省科学技术厅（GDST）（项目编号：2020B1212030009、2023A1515010757）以及香港大学（项目编号：203720465）的部分资助。

**CIM-WV数据集概览**
本研究构建了城市信息模型（City Information Model, CIM）生成的窗口景观数据集（CIM-WV），该数据集包含2000张经人工标注的图像，采集自香港高密度高层城区。
1) CIM-WV数据集的窗口景观图像覆盖香港不同区位、高程与朝向的多样化城市景观；
2) CIM-WV数据集共包含7类语义标签，分别为建筑物、天空、植被、道路、水体、车辆与地形。

此外，本项目还提供了分别基于CIM-WV数据集、真实窗口景观图像、纽约谷歌地球（Google Earth）CIM生成的窗口景观图像以及新加坡谷歌地球CIM生成的窗口景观图像训练得到的DeepLab V3+模型变体。您可在此修改源代码以调用已训练完成的DeepLab V3+模型。

**研究贡献**
1) CIM-WV是首个公开的、基于CIM生成的富语义照片级真实感窗口景观数据集；
2) 对比分析表明，相较于基于地面视角图像的深度迁移学习方法，利用CIM-WV数据集开展深度学习可实现更精准的窗口景观评估；
3) 面向城市研究人员与行业从业者，我们公开的基于CIM-WV数据集训练的深度学习模型可支撑基于多源窗口景观的新型城市应用场景，包括精准房地产估值、建成环境优化以及与窗口景观相关的城市分析。

若您的研究与实践工作中用到了本项目的成果，请引用本论文与数据集，不胜感激。
如有任何疑问，请联系Maosu（邮箱：[e-mail]）或Frank博士（邮箱：[e-mail]）。