This project proposes three optical character recognition (OCR) methodologies, where a comparison and performance analysis were carried out, all implemented in an embedded system, applied to industrial lines, whose products belong to the Snacks type classification and seek to recognize data General information such as expiration date, batch, and type of product according to company parameters. The comparison between three algorithms is proposed, one of them corresponding to an own development that corresponds to a classifier-type convolutional neural network, written in Python language. Likewise, the performance of this algorithm is compared with two open-source OCR tools for public use. A test has been developed where the recognition systems are implemented in an industrial context of a Snacks company in the industrial sector of the city of Cali, where it is achieved by means of an industrial camera, the reading of the characters and basic information of the packaging, such as lottery, expiration date, among others.