A Data-Driven Approach to Safety Mapping and Intelligent Route Planning in Lisbon

Micromobility proposes a sustainable alternative to traditional urban transportation, actively seeking to decrease emissions and promote more active lifestyles. However, ensuring the safety of these users, particularly cyclists in urban landscapes, continues to be a significant challenge. In 2019, there were 2000 cyclists killed in the European Union. To confront this issue, the current study is devoted to pinpointing unsafe points and neighborhoods, while providing the safest routes with a purposely designed route planner based on the most up-to-date Lisbon cycling network.

State-of-the-art object detection (YOLOv5), image segmentation (NVIDIA Semantic Segmentation) and an in-house developed neural network (Bikeable) were employed to predict from Google Street View images the dangerous points for cycling. And, while considering all the classified points, the safest and unsafest neighborhoods were identified in Lisbon. This unique methodology served a dual purpose: to generate comprehensive safety maps and, as importantly, to recommend personalized routes (OpenRouteService) factoring in distance, travel time, and safety.

Our findings (routeplanner.cycleai.net) underline the transformative potential of data-driven tools in reshaping urban mobility.