Skip to content

Prevention of pandemic risk

Streetscape design through artificial intelligence and design strategies

COVID-19 has become the most severe health challenge worldwide over the past years, drastically disrupting residents’ daily life. Compared with other built environments, people tend to take outdoor activities in urban streets during the pandemic. Thus, understanding risk perceptions in urban streets can promote necessary behavior changes, supporting effective public health risks management. However, traditional measurement approaches, including questionnaires and interviews, are labor-intensive, expensive, and time-consuming. In this project, we proposed an artificial intelligence-based method to rapidly and cost-effective measure the effects of the urban streets on COVID-19 risk perception.


Trained human-machine adversarial scoring models that can generate Covid-19risk perception scores

Conducted experiments related to Tianhe District streetscape assessment

Graphize the results and submit them the competition


The study was carried out in four phases: data collection, data preprocessing, data analysis, and application.
The central area of Guangzhou was an ideal study area with a high population density and COVID-19 risks.
According to the predicted results of the ResNet50 model, we mapped the spatial distribution of COVID-19 risk perceptions.
Based on the regression results, we analyzed the potential mechanisms and proposed the overall planning strategies.
Responding to dynamic characteristics of risk perceptions, design strategies were proposed based on regression results and potential mechanisms.
For the narrow pedestrian space lacking greenery, we created resilient residential streets to optimize the COVID-19 risk perceptions.
For outdoor dining areas and disorderly queues, we reduced overoptimism perceptions and the risks of COVID-19 infection.
Reducing enclosure and increasing scene diversity provide essential and comfortable outdoor respites for residents in leisure streets.
We evaluated the design proposal to test the effectiveness of risk communication on the human-machine adversarial scoring platform.