Smart Billing Cart using Deep Learning for Mall Administration

Abstract

Customers go to the mall to purchase and spend money on the goods they require. Before generating a customer's bill, it is preferable to confirm how many items have been sold out. One must make an effort to select the right item while going to the store to make a purchase. After that, waiting in queue to get invoiced is stressful. Therefore, we propose a smart cart solution to record both online invoicing transactions and purchased goods without using radiofrequency identification (RFID). The technology will also provide product recommendations based on user purchasing patterns from a centralized source. In this system, each cart has a camera attached to it, and each item in the shop will have identified using YOLOv4. The transaction and recommendation systems for the web will be centralized. The recommended system consists of a camera that weighs the attached object and a load cell that utilizes deep learning to determine the item. The system will generate the bill when the consumer scans an item with the help of the cart’s fixed camera. Numerous methods may be used to accomplish object recognition Bounding boxes are produced by techniques like region-based convolution neural network (R-CNN) using area proposals, and these boxes are utilized to operate classifiers continuously. After that, the duplicates are eliminated using a post-processing technique. R-CNN is a slow method of object recognition. We do this by using the YOLO concept.