Digital Future Workshop 2020
Imaging Landscape: Computer Vision and Landscape Perception
Design Workshop @ Tongji University
Co-instructed with Prof. Brad Cantrell (UVA) & Xun LIU (Ph.D. UVA)
The research on the visual impact of the human perceived landscape is limited by the lack of quantitative methods on landscape perceptions or difficulty in dealing with large amounts of image data. This one-week workshop applied computer vision and machine learning to develop an effective approach to quantify the subjective perception of landscapes and to apply to regional-scale studies. Taking Berlin as a case study, it examined the spatial formation, landscape elements composition, and landscape perception through data mining and data visualization of urban image data.
The goal was to challenge the conventional perceptual study in the landscape design process. We hope this workshop could inspire more possibilities for novel quantitative analysis and an evidence-based design approach. This one-week workshop included a four-day session with theoretical lectures and technical workshops, handing the students basic knowledge of google street view API, SVI downloading, data preprocessing, Computer Vision( pspNet, Mask RCNN), and basic Machine Learning. Then a three-day working session allowed students to develop their own projects in sub-groups. The students included undergrads, graduate students, Ph.D. students, and lecturers from different backgrounds in landscape architecture, architecture, and urban planning from China and the U.S.
Project 1: Berlin Living Environment Recommendation.
“If you are moving across the country and looking for a new home in Berlin, we are here to help. Our website will guide you through to find places that best match your preference and personal lifestyle. ”
Dan Luo ( U.Queensland / Lecturer, Tsinghua U./PhD, GSAPP/Master, HKU/B.A.)
Peng Chen (Tongji U./M.L.A student)
Leshan Fu (Southeast University/B.L.A. student)
Yuwen Yang (HKU/Ph.D. student, U.Georgia/MLA, U.Maine/BS)
Project 2: Urban Summer, Distanced
This project explores ways Berlin's public spaces could adapt to the SARS-COV-2 pandemic with social distancing measures. When people are advised to stay 1.5 meters apart from each other, streetscape dimensions and gathering patterns that used to be deemed ideal for urban living pose risks for public health. To mitigate the demand for outdoor activities in the summer and the need for maintaining safe social distances, we propose tactical recommendations for Berlin's streets and squares. The proposals are derived from comparative studies with city-wide GIS analysis of sidewalk width and points of interest, as well as computer visioning of pre-COVID Google Street Views and post-COVID webcam footage from some of Berlin's popular public spaces.
Di Zeng (Harvard GSD/MDes in CC, Tongji U./B.E. in Urban Planning)
Xin Feng (Harvard GSD/ MLA ap + MDes Tech)
Shaun Wu (Harvard GSD/MLA, Cornell AAP/B.Arch.)
Project3：Re-discovering East Berlin – Mapping of Urban Fragments under Reunification
We identify and analyze “east berlin” moments that they preserve and celebrate in the reunified city through mapping. Through trial and error, we applied different combinations of the data matrix to Berlin’s central area, in order to study if specific patterns could help us find leftover fragments of East Berlin.
Hui Tian (Louisiana State U./ MLA, Tianjin U./ MLA, China Agricultural University/ BLA)
Weishun Xu (Zhejiang U./Lecture, GSD/ MArch 15’, UVa/ BArch)
Ziyu Han (Tongji U/ B.E. in Urban Planning)
Project 4: Childhood Obesity in Berlin | A Digital Urban Study
Childhood obesity in Berlin has been increasingly cited as a major health issue in recent years. In this research, we will identify significant effects regarding the social index, accessibility to open space and street quality data, and spatially visualize their correlations to this challenge from an individual to a neighborhood level.
Lide Li (The University of Cambridge/ 2020Fall master student, South China University of Technology/ Bachelor)
Huiwen Shi (MIT SMARCHS Urbanism/ 2020Fall master student, South China University of Technology/ Bachelor degree）
Project 5: Urban screening: What spatial features influence outdoor sports choice
In urban environments, people tend to use the streets that make them comfortable. The feeling is aroused through various spatial elements. In this study, we are exploring the value of each spatial element by comparing the percentage of spatial elements and exercise heat map. Computer vision techniques including Semantic Segmentation and Mask RCNN are used to support our research process. At the end of the research, we can speculate about the relationship between spatial elements and people’s preference for outdoor exercise on various streets.
Chun Xu (Cornell AAP/2020Fall MRP student, Tongji University/BArch)
Yuechen Cui (Beijing University of Technology/Current Undergrad)
Austin Lu (GSD/ current MLA student, Louisiana State University/ BLA)