Realtime Customer Merchandise Engagement Detection and Customer Attribute Estimation with Edge Device

In this paper, a low-cost computer vision-based system has been developed which detects and monitors the interaction between customer and merchandise. Various computer vision technologies such as real-time object detection, object tracking and person identification have been used in this work. The proposed system has been divided into two subsystems. First, the customer and merchandise interaction detection system while second, the age-gender estimation server. The two subsystems establish their communication through TCP socket. Through extensive evaluation and experiments, our proposed lightweight and low-cost system successfully achieves high accuracy in detecting the interaction in real time.