Deep Urban Analytics: Predicting Gender-specific Urban Safety using Artificial Intelligence
Image source: Course Faculty

Urban safety perception has long been studied through active measures such as patrolling, lighting at night, and grievance addressal systems. The use of machine vision allows designers to interrogate how urban fabric explicitly affects spatial perception. This course is designed to be an introduction to the use of artificial intelligence in addressing the age-old question of “How to quantify and predict subjective spatial experience?”

Participants will be introduced to the concepts of artificial intelligence. As part of the exercise, semantic segmentation, object detection and colour detection will be used to extract and quantify urban features (e.g., people, trees, vehicles, sky, buildings, colour distribution, etc.) of photographs of urban scenes. Subsequently, participants will rate the photographs with respect to safety perceptions. The corresponding quantified urban features and ratings will be used to train a neural network to predict the perceived safety of new urban scenes. Finally, with the aid of the trained neural network, participants will edit features of new urban scenes (as an exercise of design intervention) to increase or decrease perceived safety.

Course Faculty

Student Projects