Chen Yi's team at Future Cities Lab Global built an urban search engine that uses AI to analyze street-level imagery and extract visual elements like trees, benches, and buildings. Designers can query the system to find real-world examples matching specific characteristics, like open sky or high vegetation, turning it into an empirical tool for testing design assumptions.
Chapters
Foreign. Hi, my name is Chen Ni Cai. I'm currently a postdoc researcher at Semantic Urban Elements Module in Future Cities Lab Global in Singapore. My work focuses on computational urban design, which is the intersection of AI computation and urban design. So I'm working towards developing design instruments to bridge city science and urban design. Because science and design, they evolve together. I'm currently focusing on street environment. Imagine that if you walk down on the street, elements such as roads, benches, trees and so on, all of them come together and shape the unique place and also shape the unique working experience. So one part of my work is to understand how design and the combination of those elements can affect the experience of different groups of people. And just as important is that we need to bring and turn this understanding and evidence to inform better design decisions. In one of my ongoing research projects, we are developing urban search engine. And it is not only about search. It is actually an automated pipeline collecting data from all across the city, analyze them and in the end allowing users to search for places based on the urban characteristics that are defined by themselves. We are using techniques such as AI to extract the visual elements, for instance roads, vegetation, trees, buildings and so on from the street view images. And then we built information retrieval workflow, so then the users can explore the real world examples in their cities based on the urban characteristics that they want to look for. So it's somehow like checking through dictionaries, but it's urban dictionaries and the alphabets are actually the urban characteristics in the built environment. So for instance, you can search for places that have more open sky or places full of vegetation, for example. As urban designers, while you have certain assumptions of what characteristics contribute to what kind of urban design qualities, using the search engine based on those cases, you would be either to prove your assumptions or contradict them. So using the search engine, we can inform better design decision making. While machines are powerful, we need to bring human in the loop. So using the design instruments that we developed, people can really define and interact with and help validate what they search for. The tool is not meant to give optimal solutions for everyone, but to help human urban designers to develop design strategies to meet the different needs of user types.
