Aprendizado profundo para classificação de microorganismos

Abstract

In this work we present a comparison among three machine learning (ML) methods applied to identify Chattonella based on microscopic images. We analyze KNearest Neighbors algorithm (KNN), Weightless Neural Network algorithm (WNN) and Convolutional Neural Network algorithm (CNN). The latter being the state of art to image classification. The goal of this work is to identify the best method to count different types of microorganisms in oceanic water samples, with the goal for detecting pollution. This comparison takes into account accuracy of the hit rate. The best result was reached by Convolutional Neural Network algorithm, as will be described in this work.