A Transfer Learning Based Techniques: Vegetables Disease Recognition and Classification

Abstract

A worldwide palatable food is vegetables cultivated by the farmer and it is most important for a healthy body. There are three different vegetables such as Malabar leaves, Bean Leaves and Papaya Leaves and fruits have been used in this approach because of the big source of vitamins, calcium, iron etc. Total 887 vegetables data have been used for classifying between disease affected and Healthy vegetables using most transfer learning techniques such as MobileNet-V2, InceptionResNet-V2, Deep CNN, NasNetLarge and ResNet152-V2 pre-trained model. InceptionResNet-V2 performs better than others applied pretrained models with 93.60% accuracy with the lowest loss value and there have been multiple activation functions as elu and ReLu, and for the dense layer mainly used sigmoid and softmax function for classification of the model. The proposed approach is a more acceptable model compared to existing models based on the accuracy, recall, precision and f1 Score.