Enhancing Retail Checkout Efficiency Through a Hybrid YOLOv8-Based Grocery Detection and Billing System

Abstract

Traditional retail checkout systems that rely on barcodes are limited at handling the non barcoded shopping items like fresh produce. To tackle this problem, this paper will introduce a deep learning approach to grocery item detection and automated billing system through computer vision. A local dataset comprising 1,109 images in 27 grocery classes was gathered and labelled and augmented with Roboflow to enhance stability. Several object detection models based on the YOLO algorithm such as, YOLOv5s, YOLOv8n, YOLOv8s, YOLOv11n and YOLOv12n were trained and tested in unvaried conditions. Experimental outcomes demonstrate that more recent architectures are far superior in detecting, with YOLOv12n having the best localization accuracy on the test set (mAP@0.5:0.95 of 0.847), and the YOLOv8s giving good tradeoffs between accuracy and efficiency to be used in the real world. The chosen model (YOLOv8s) was incorporated into a web-based image-based billing system, which confirmed the possibility of a scalable and inexpensive AI-based checkout system. The suggested framework provides a platform upon which future research can proceed to enhance the real-time performance and dataset generalization.