Dr. Matias Quintana describes a dataset of 10 million images across 688 cities used to measure urban qualities like greenery and walkability. The team trained machine learning models and collected human perception rankings to build AI that informs urban design for different demographics.
Chapters
Hi, My name is Dr. Matias Quintana. I'm a postdoctoral researcher and module coordinator here at Future Cities Lab Global in Singapore. My current research is looking at cities through the lens of images. Images, for example, studio images are these images that you may have seen cars with a camera on top collecting through the streets. We can compute a lot of different things. How much greenery is in there, whether it's a crime, maybe. In collaboration with NUS, we have collected this massive data set of 688 cities around the globe. We have more than 10 million distributed images with more than 300 metadata like features, like columns on a table for each image. These allow us to have these sort of labels that we can then compute into numeric indicators to allow us to train machine learning models that will serve to predict an image that was not seen before. We have conducted another experiment of giving these images to people and allowing them to rate them, meaning they will see two images and compare them under a given question and say which one is more safe, which one is more beautiful. We have found that there is a lot of correlations between the age of a citizen and how they perceive a specific environment. How safety will appear to them based on what's in the street, the more greenery, more accessible walking spaces, like the width of the walking lane. This is useful because it allows us to then give guidelines or design practices to urban planners and designers to better address needs of its current citizens based on demographics. These data set allow us to have this human centricness. I think we need human centricness in AI because not only it's useful for diversity, we are an entity that looks at global problems. Even though we're based in Singapore and Zurich, we need this focus on humans as cities are developed and built by and for for that.
