← Work
Machine Learning & Perception
Traffic Inefficiency Identification via Computer Vision
Cut traffic-data-collection labor by 80% with a YOLOv8 pipeline, after consulting Hawaii's DOT.
YOLOv8 with Microsoft Mosaic Imaging detects and tracks vehicles across stitched high-resolution intersection composites, replacing manual counting over six months of field testing.
It’s paired with a novel traffic-efficiency model that drops the time hyperparameter and optimizes to minimize interquartile range while maximizing efficiency — developed after consulting Hawaii’s Department of Transportation.
