Microbial Image Deciphering: Navigating Challenges with Machine and Deep Learning

Abstract

Abstract This paper presents a novel approach to microorganism classification through the use of Convolutional Neural Networks (CNNs), demonstrating the potent capabilities of deep learning in the realm of microscopic image analysis. Utilizing a rich dataset of microorganism imagery, captured with a Canon EOS 250d Camera and meticulously categorized into eight distinct classes, we have trained a sequential CNN model that effectively distinguishes between various microorganisms with high precision. The dataset, comprising images in JPEG format, was sourced from the controlled environment of Pathantula Tea Garden's laboratory settings, ensuring consistency and quality in data acquisition. The CNN architecture, designed with layers of convolution, max pooling, and dense operations, further refined with dropout and batch normalization, has been optimized with several optimizers including SGD, RMSprop, Adam, and Nadam, all set at a learning rate of 0.001. Notably, the Adam optimizer emerged superior, propelling the model to achieve an impressive 97% accuracy. This research not only underscores the efficacy of CNNs in classifying microorganisms but also paves the way for future advancements in automated microscopic image classification.