Comparative Study of Machine Learning and Deep Learning for Fungi Classification

Abstract

The aim of this paper is to compare and contrast between deep learning (DL) and various machine learning (ML) algorithms for fungi classification. The Danish Fungi data set provided by Kaggle, for this study. Only, 10 classes from the provided data set were extracted which consists of 1775 images. In this work, the used machine learning techniques are decision tree (DT), Naive Bayes (NB), K-nearest neighbour (KNN) and random forest tree (RFT) and achieved accuracies of 25, 28, 29 and 33%, respectively. The reason for low accuracies for the machine learning algorithms is because machine learning algorithms are usually used for numerical data and not suitable for images. Deep learning model using Keras was used to achieve an accuracy of 75.82%. On comparing the quantitative metrics like precision, recall, f1-scores, it can be concluded that deep learning algorithms are much better than machine learning algorithms.