Visual-based People Counting and Profiling System for Use in Retail Data Analytics

Abstract

Data on various key performance indicators (KPIs) are crucial in preventing problems and growing a business. In this paper, we propose the creation and analysis of the feasibility of using an intelligent video analytics (IVA) system to gather data on certain restaurant key performance indicators (KPIs). The main challenge lies in maximizing the use of an existing CCTV camera with a fixed viewpoint, which is tailored for security purposes instead of video analytics, by using its footage in the IVA. The researchers partnered with a restaurant in a high-traffic business district to create and test the system. The final system gathered data on foot traffic, customer gender classification, and customer group size. Neural networks such as YOLO, Deep SORT, and InceptionV3 were employed in the implementation. The results show that while it is possible to gather data on these three metrics through the system, the speed and accuracy can still be improved through downsizing the frames, down sampling the videos, and using other algorithms.