A smart scanner system for ingredient categorization and identification of nutritional composition in packaged food items

Abstract

Food constitutes a vital part of the human lifestyle. Packaged food is not avoidable, especially for working professionals. These are highly processed and are sometimes high in artificial colors, preservatives, saturated fats, sodium, and other unhealthy ingredients. It is necessary to identify and avoid the intake of allergic and unhealthy ingredients in packaged food.  This research introduces a pioneering AI-based application that seamlessly integrates Optical Character Recognition (OCR) technology with an extensive database of ingredient information, offering users detailed insights into the nutritional composition, categorization, and potential allergens of diverse food products. Going beyond conventional models, this application employs OCR for text extraction, empowering users to scan a wide array of food items beyond those with barcodes. The study evaluates the accuracy of text extraction and the performance of an innovative ingredient categorization model. Google's Vision API emerges as the optimal choice for text extraction, demonstrating exceptional results. The categorization model achieves an accuracy of 84% and a precision of 87%, providing users with a reliable tool to make informed decisions about their dietary choices, ultimately contributing to enhanced health and well-being.