Comparative evaluation of deep learning models for Yellow Mosaic Disease identification in blackgram

Abstract

In recent times deep learning has received a lot of attention for image classification and identification. It is widely being used for plant leaf disease identification and classification using different models. The present study aims at assessing different pre-trained deep learning models viz., AlexNet, GoogLeNet, and ResNet-50 for the classification of images for Yellow Mosaic Disease (YMD) in blackgram. Datasets consisting of images of three classes namely healthy, moderate and susceptible were collected during the field study. Images were pre-processed and augmented to make a large dataset for the training purpose. A total of 70% of the images were used for the model training and 30% were used for the validation and testing. Validation accuracy and loss for different models namely AlexNet, GoogLeNet, and ResNet-50 showed values of 95.39, 95.86, and 94.83 and 0.3544, 0.1365, and 0.1317 respectively. All the models worked well for the YMD disease classification in blackgram but AlexNet took the smallest computational time and ResNet-50 took the longest computational time.