Authors:
Dang Thi Phuc
1
;
Dau Sy Hieu
2
;
Nguyen Manh Hoang
1
and
Tran Thi Minh Khoa
1
Affiliations:
1
Department of Computer Science, Faculty of Information Technology, Industrial University of Ho Chi Minh City, Ho Chi Minh City, Vietnam
;
2
Department of Applied Physics, University of Technology, Vietnam National University HCMC, Ho Chi Minh City, Vietnam
Keyword(s):
Internet of Thing, Artificial Intelligence, YOLOv4, Embedded System, Jetson Nano.
Abstract:
Shopping in shopping malls and supermarket is gradually increasing and replacing traditional market because of the convenient conditions such as full of products, clean place with all time air-conditioner, modern environment, etc. Supermarket owner always want to find the way that attract more and more customers come to the supermarket as well as advertise their products to the customers as many as possible. In order to offer relevant and attractive advertising to customers, detection and classification customers entering to the supermarket is taken into consideration. Due to the characteristics different customer groups, the relevant products should be showed to attract customer, help them save the time for finding products. In this paper, we build a real-time customer detection and classification system at the supermarket. The goal of this proposed Internet of Things (IoT) system is automatically show the suitable advertising clips to many customers at the right time. We build a cl
assification model using deep learning with a large amount of data. The dataset is collected from reality and labelled with five different object classes. To ensure reliability, 7000 images are collected from different conditions such as variations in camera used, bad lighting, angles, and not stable background. The data is trained on YOLOv4 and YOLOv4-tiny models. The models are deployed on the embedded system with the Jetson Nano device as the processor. We compare the accuracy and speed of the two models on the same embedded system, analyse the results, and chose the best model according to the specific system requirements.
(More)