AutoLTS: Automating Cycling Stress Assessment via Contrastive Learning and Spatial Post-processing

Cycling stress assessment, which quantifies cyclists' perceived stress imposed by the built environment and motor traffics, increasingly informs cycling infrastructure planning and cycling route recommendation. However, currently calculating cycling stress is slow and data-intensive, which hinders its broader application. In this paper, We propose a deep learning framework to support accurate, fast, and large-scale cycling stress assessments for urban road networks based on street-view images. Our framework features i) a contrastive learning approach that leverages the ordinal relationship among cycling stress labels, and ii) a post-processing technique that enforces spatial smoothness into our predictions. On a dataset of 39,153 road segments collected in Toronto, Canada, our results demonstrate the effectiveness of our deep learning framework and the value of using image data for cycling stress assessment in the absence of highquality road geometry and motor traffic data. * If one lane per direction and motor traffic speed ≤ 40 km/h, then LTS 1. * If one lane per direction and motor traffic speed ≤ 48 km/h, then LTS 2. * If motor traffic speed ≤ 56 km/h, then LTS 3. * Otherwise, LTS4. -If the road segment has no on-street parking, * If one lane per direction and motor traffic speed ≤ 48 km/h, then LTS 1. * If one/two lanes per direction, then LTS 2. * If motor traffic speed ≤ 56 km/h, then LTS 3 * Otherwise, LTS 4. • For road segments without cycling infrastructure: -If motor traffic speed ≤ 40 km/h, and ≤ 3 lanes in both directions, * If daily motor traffc volume ≤ 3000, then LTS 1. * Otherwise, LTS 2. -If motor traffic speed ≤ 48 km/h, and ≤ 3 lanes in both directions, * If daily motor traffc volume ≤ 3000, then LTS 2. * Otherwise, LTS 3. -If motor traffic speed ≤ 40 km/h, and ≤ 5 lanes in both directions, then LTS 3. -Otherwise, LTS 4.