Automatic detection and classification of brewer's yeast for viability analysis

Abstract

The viability and vitality of the yeast Saccharomyces cerevisiae are crucial for brewing. The traditional method for assessing viability, manual counting with methylene blue, is laborious and prone to variability depending on the individual counting. This work addresses the development of an algorithm for the automatic counting and classification of brewing yeast from microscope images using artificial intelligence tools, with the aim of optimizing the reuse and quality control of this input for the industry. The proposed system integrates two main stages: cell detection is performed using the YOLO (You Only Look Once) neural network, while classification into live or dead cells is carried out using a Support Vector Machine (SVM) based on saturation channel histograms in the HSV color space. This combined approach offers a robust and automated solution for yeast quantification and classification, reducing the variability of manual analysis and optimizing quality control in brewing production. The developed methodologies will be incorporated into the Microbrew.AR application.