Identifying customer interest from surveillance camera based on deep learning

Abstract

This study proposes a method to identify a customer's interest in the product. Specifically, we applied the state-of-the-art deep learning algorithms to the real-world surveillance videos for analyzing customer interest in the product and evaluated the accuracy. For this, we first introduce a new first of its kind dataset called ICI (items of customer's interest) that includes various shopping situations. We experimented the state-of-the-art deep learning algorithms on the ICI dataset to determine a suitable algorithm for identifying a customer's interest. The experimental results demonstrated that the estimation accuracy is 71% on the average, meaning that a customer's interest can be measured effectively. © 2020 IEEE.