Abstract:

Retail shelf planning is a critical component in optimizing product placement and enhancing the overall shopping experience for customers while simultaneously addressing the challenge of reducing excess inventory that ties up capital. In recent years, the integration of image analytics techniques has brought about a transformative shift in the field of retail shelf planning by offering automated, data-driven solutions.

This paper presents a practical application of cutting-edge object detection techniques in the context of retail shelf management. Visual audits are conducted collaboratively by retail staff and representatives from Fast-Moving Consumer Goods (FMCG) product companies, whose products are stocked on these shelves, with the primary aim of identifying out-of-stock products. The central objective of this paper is to address the challenge of “Classify objects into recognized classes and count instances,” a long-standing issue in the field. Successfully solving this problem has the potential to optimize operational efficiency by reallocating human resources to areas where their expertise surpasses that of machines. In this study, the You Only Look Once (YOLO) V8 nano model is employed and trained to detect two specific object classes relevant to the retail shelf environment. The results showcase a precision mean Average Precision at 50% intersection over union (mAP50) of 0.871 across all object classes, while the mAP50-95 stands at 0.541, indicating robust object detection capabilities. This paper concludes with a forward-looking discussion on the future directions for this paper and underscores the substantial impact that image analytics is poised to make in shaping the future landscape of retail shelf planning. The convergence of advanced technology and retail industry expertise promises to further optimize product availability and customer satisfaction while streamlining operational processes.

Keywords: Object Detection, Neural Networks, Image Analytics, YOLO V8.

Conference Name: 2023 Global Conference on Information Technologies and Communications (GCITC)