Abstract
Currently, the shelves of shops and supermarkets are filled with food that people consume daily, with many products coming from abroad. However, are all these products useful for the human body, and do they meet the standards? In this article, we will talk about how to identify low-quality products using modern machine learning. Recognition and classification of images and text based on machine learning can be a key technology in the fight against low[1]quality food. Automatic image and text recognition and classification of product information enable end customers to identify counterfeit products accurately and quickly by comparing them to trained templates. However, it is clear that this does not apply to all food processing enterprises. In food production, low-quality and non-standard products are used to reduce the cost of the product. Manufacturers can change their products by replacing higher quality products with lower quality ones. They may use confusing terms on the label to mislead you. When buying and serving counterfeit products, consumers suffer in different ways. First, they may not be getting the nutrients they need, adulterated foods may not be safe for their health, and may also be an economic loss for consumers. We evaluate the technical feasibility of the components of this food fraud detection architecture using a real-world scenario, including machine learning models to distinguish multiple products from each other. It allows you to control the circulation of food products at the state level, thereby protecting the end consumer from purchasing low-quality and potentially dangerous goods. In this article, we used the MobileNetV2 model and multiclass classification and evaluated the model we received from different angles.