Performance Evaluation of Deep Learning Models for Leaf Disease Detection: A Comparative Study

Abstract

Early detection of plant diseases is crucial before plant growth is affected. Plant diseases have been detected and classified using a variety of machine learning (ML) models in the past. Deep Learning (DL) appears to have great potential in terms of increased accuracy; however, in agricultural applications of Convolutional Neural Networks (CNN) has widely been utilised by researchers. CNNs are so effective at identifying plant species, managing yields, detecting weeds, managing soil, and water, counting fruits, detecting diseases and pests, and evaluating plant nutrient status. A farmer can diagnose plant diseases quickly and accurately with an automated disease detection system. To speed up crop diagnosis, plant leaf disease detection systems must be automated. In this paper, we evaluated twelve different models on a new plant diseases dataset and demonstrated that the most accurate model was Densenet169. In training and validation, the accuracy was 97.2% and 97.8%, respectively.