Pedestrian Trajectories and Thermal Exposure Data
Description
This dataset comprises four components supporting the analysis of pedestrian walking behavior under outdoor heat exposure. The first component contains a three-dimensional model of the experimental environment reconstructed from UAV oblique photogrammetry, providing the spatial reference for scene representation and trajectory georeferencing. The second component contains the code used for pedestrian detection, multi-object tracking, and trajectory extraction based on YOLOv8-SPD and DeepSORT. The third component provides the MASt3R matching data, including image-to-scene correspondences and control points used to transform pedestrian trajectories from image coordinates into metric real-world coordinates. The fourth component contains the ENVI-met simulation outputs for Experiments A and B in a single folder. These outputs provide the pedestrian-scale microclimate conditions and thermal-comfort indicators used to examine pedestrian responses to shade and heat exposure.
Files
Steps to reproduce
1. Download and extract all folders, then install the required Python packages and model dependencies specified in the code documentation. 2. Prepare a pedestrian video, configure the input path and model weights, and run the YOLOv8-SPD and DeepSORT pipeline to obtain frame-level pedestrian detections and image-coordinate trajectories. Raw videos from the study are not provided because of privacy considerations. 3. Use the UAV-derived three-dimensional scene and the provided MASt3R control-point correspondences to estimate the spatial transformation and convert the image-coordinate trajectories into metric local coordinates. 4. Load the ENVI-met outputs for Experiments A and B, extract the relevant thermal variables, and associate them with the transformed trajectories according to location and observation time. The resulting data can be used to calculate walking speed, thermal exposure, and route-choice indicators.