Annotation: The problem of distortion and substitution of information when performing transactions in retail trade using self-service checkouts leads to a decrease in the profits of enterprises. The introduction of weight recognition tools using machine-learning methods will increase the level of protection for retail transactions. The paper analyzes various training methods and builds a network model that allows, with a small amount of labeled data, to obtain good recognition results in the field of retail.

Keywords: convolutional neural networks, machine learning, retail information security