کنفرانس مهندسی برق ایران

صفحه اصلی / سی و چهارمین کنفرانس بین المللی مهندسی برق

A Low-Cost System for Real-Time Retail Product Recognition on Embedded Platforms

نویسندگان :

Hossein Haghjoo¹ Seyed Ahmad Khalilpour² HamidReza Taghirad³

1- دانشگاه تهران، دانشکده مهندسی برق و کامپیوتر 2- دانشگاه صنعتی خواجه نصیرالدین طوسی 3- دانشگاه صنعتی خواجه نصیرالدین طوسی

کلمات کلیدی :

Smart Store،Embedded Systems،Raspberry Pi،Deep Learning،Machine Vision،Dataset Creation

چکیده :

Abstract—Nowadays, with increasing competition in the retail industry, the need for intelligent and automated man- agement systems to increase eﬀiciency and reduce human error is felt more than ever. This project details the design and implementation of an intelligent system for managing stores using modern machine vision and deep learning. A primary challenge was the lack of public and suitable image datasets for local retail goods. To solve this, a dedicated hardware system with three cameras, a rotating plate, and an Arduino controller was designed and built for automated product imaging, enabling the creation of a unique dataset. Following this, various deep learning models were trained for accurate product detection. A key challenge, the optimization and eﬀicient implementation of the most effective model on the limited hardware of a Raspberry Pi, was successfully accomplished to achieve low-latency performance suitable for interactive checkouts. The main achievement is an integrated solution that deploys this optimized model on a Raspberry Pi for instant product recognition. In addition, an advanced management panel was developed using the Django framework and Python. This integrated system is a significant step towards the complete smart automation of store processes.