Decoding DOOH viewability using YOLO for privacy-friendly human silhouette identification on LiDAR point clouds

Abstract

Traditional methods for measuring digital out-of-home (DOOH) advertising effectiveness often rely on static data or camera footage, leading to limitations in accuracy and real-time insights. This research proposes a novel approach that leverages the combined power of LiDAR technology and YOLOv8, a state-of-the-art object detection model, to achieve precise and privacy-friendly human silhouette identification for DOOH performance measurement. By extracting 3D point cloud data from LiDAR sensors and employing YOLOv8's efficient object detection capabilities, the model accurately identifies and tracks pedestrians in the vicinity of DOOH displays. This information, combined with LiDAR's performance under varying weather and lighting conditions, offers a significant improvement over traditional methods, providing advertisers with valuable real-time data on audience engagement and campaign effectiveness. The comparison with the same model performance trained on a standard MC-COCO 2017 dataset presented comparable accuracy but faster inference times. Furthermore, the focus on LiDAR data ensures privacy by avoiding the use of facial recognition or other sensitive personal information. This research demonstrates the feasibility and potential of LiDAR-based human silhouette identification for DOOH performance measurement, paving the way for a more data-driven and effective advertising landscape.