Design and Evaluation of a Real-Time Stock Inventory Management System

An efficient stock inventory management is a vital part of a company, stores, laboratory, hospitals etc to balance stock with the level of demand and reduce the amount of manpower and money required for manual management. In addition to this, safety of the inventory also becomes important to reduce any risks of theft leading to potential loss to the owners. In this paper, design of a real-time stock inventory management system is proposed. An application is also developed for the end user to monitor and access the real-time inventory management system. The proposed system is implemented using three machine learning models - You Only Look Once (YOLO), Single Shot Detector (SSD) and Faster Region-based Convolutional Neural Network (Faster-RCNN) and their performances are compared in the proposed work. The performance evaluation of the proposed system is carried out using real-time videos captured from a laptop camera. The proposed system yielded an average precision of 0.92, 0.893 and 0.89 with an average recognition time of 1.33s, 0.33s, and 1.81s on using YOLO, SSD, and Faster-RCNN algorithm respectively. The proposed system also allows the end user to stream the real-time video remotely along with the details of objects detected and notifies the user immediately on detecting an intruder in the environment. The proposed system can be employed in showrooms for electronic gadgets, hypermarkets and retail stores to track the stock and detection of intruders at odd times.