- Cases
- Netherlands
Hackathon: Mapping Accessibility in Rotterdam with Artificial Intelligence developed by uni students.
01 Situation
Rotterdam’s municipality sought to identify mobility barriers in public spaces—such as scooters on sidewalks or missing ramps—to improve accessibility for residents and visitors with mobility constraints.
02 Assignment
Bachelor and master students were tasked with gathering and labelling up to 10,000 street-level images indicating “accessible” or “not accessible” zones. Their goal: create a training dataset for an AI system capable of recognizing obstacles in urban environments.
03 Approach
Working in groups of four, students captured images across diverse neighbourhoods, followed by critical documentation of their collection methods, choices, and ethical considerations. Each student also contributed an individual reflection exploring the social impact of algorithmic decisions and the role of data in shaping equitable urban futures.
04 Result
Datasets were evaluated by training an Inception v3 model on each submission, revealing which group’s data best reduced bias while improving accuracy on a diverse test set. Students gained experiential insight into data-driven design while directly contributing to Rotterdam’s accessibility efforts.
Enjoyed this case and want to dive deeper? We will connect you with the designers who worked on it.