Transfer Learning Using Convolutional Neural Network to Classify Leaf Diseases on Ridge Gourd Plant

Leaf disease can have a long-term effect on vegetable growth. If proper care is not taken, it may have adverse effects on vegetable production. The disease on the leaf should be identified on time to avoid further deterioration and reduction in the yield of plants. In this study, we compare various image classification models created using Convolutional Neural Networks for the leaf diseases classification of a ridge gourd plant. Along with healthy leaf images, the dataset contains images of leaves affected by Larvae, Anthracnose, and Downy Mildew. Pre-trained image classification models like VGG16, VGG19, ResNet50, and MobileNetV1 are compared. Best fitted CNN model is designed by tuning the hyperparameters namely the number of filters, kernel size, activation function, and the optimizer. Results obtained with the proposed CNN model are compared against pre-trained models based on accuracy and training time. Results show that the proposed CNN model takes less training time and gives better accuracy compared to other pre-trained models.