Automated identification of soil Fungi and Chromista through Convolutional Neural Networks

The identification of soil microorganisms plays a crucial role in agriculture and horticulture, as it enables the monitoring of beneficial species and early detection of pathogens. In this study, we propose a system that utilizes machine vision and machine learning techniques, specifically Convolutional Neural Networks, to automate the identification of different fungi and Chromista based on microscopic images and morphological traits. Our system aims to provide a cost-effective and efficient method for pathogen detection, improving the overall health and productivity of agricultural systems. We conducted experiments using a dataset of soil microorganisms and the performance of the classifier was evaluated using precision, recall and F1-score measures. Despite challenges such as class imbalance and imperfect subimage retrieval, the classifier achieved promising results, with an overall precision of 82% indicating the high accuracy of correctly predicted positive instances across all classes. Furthermore, the incorporation of a majority voting scheme significantly improved the classifier's performance, addressing the issue of underrepresented classes. The enhanced results demonstrated an average precision and F1-score of 97%. Our work highlights the potential of CNNs in soil microorganism identification and paves the way for future research to expand the dataset and to incorporate a wider range of microorganism genera.