This codebase is in support of FARLAB's XCPedestrian project, where we seek to quantify differences in pedestrian-vehicle interactions across different locales.
This contains code for CV processing -- pipeline.py is the main script we use to (1) run COCO-trained YOLO on urban footage, returning instance counts of 80 classes; including pedestrians, cars, and bikes; and (2) isolate relevant clips that contain intersections for downstream qualitative coding.
This contains SLURM job scripts for triggering the scripts in an HPC environment. Modify these as needed for your own system.