This webpage is designed to showcase the cyclist stress heatmap, a key outcome of our proposed cyclist stress assessment model, rather than to display the entire paper/project. Given the extensive size of the heatmap, it cannot be adequately accommodated within the confines of a 12-page submission.

Project Description

Smartphone-based Cyclist Stress Assessment: To capture cyclist 'in-the-moment' stress, we develop CycStress, a novel smartphone-based cyclist stress assessment model leveraging context sensing. The model is built upon a combination of a Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM). Since obtaining a well-annotated dataset in cyclist stress presents significant challenges due to safety concerns, we create the annotated dataset on a bike simulator and leverage a contrastive learning framework to bridge the simulation-to-reality gap. In practice, CycStress only activates a minimal sensor set to avoid excessive energy consumption of mobile devices. In the video below, we show how CycStress works in real-world scenarios.

Cyclist Stress Heatmap: We provide a stress distribution heat map for the study area, representing an aggregate of stress assessments from all participants. This map serves as a valuable tool for city planners, enabling them to pinpoint areas requiring enhancements in cycling infrastructure. Additionally, it allows cyclists to select routes with minimal stress or even stress-free options for their daily commutes.

Note: The above image is a screenshot from our produced interactive heatmap, which showcases the accumulated distribution of cyclist stress. Upon acceptance of the paper, we plan to replace this static image with the original interactive map, featuring enhanced functionalities like zoom-in and zoom-out options. The use of a static map in this submission adheres to the conference's requirements for anonymization.