Classification of Multi Plant Leaf Diseases Based on Optimization of the Convolutional Neural Network Models

Early identification of plant diseases reduces the adverse effects on crops. Convolutional neural networks, particularly deep learning, are frequently used in computer vision and recognition of pattern tasks. In this paper, we proposed a method for classifying multi-plant diseases of leaves using convolutional neural network models that were optimized. The proposed approach uses a classification with multiple labels approach that simultaneously classify the plant leaves and the disease type, despite using multi-class classification. As a result, we adopted many CNN models, including AlexNet, Inception V3, GoogLeNet, ResNet 18, and ResNet 50. Particle Swarm Optimization guides these pre-trained models to get better performance. PlantVillage dataset to train and test the models was used in all experiments, including many plant diseases and over 31,000 images divided into 21 disease groups of 5 plant species, including safe leaf images. The photographs were taken in an unregulated environment and with their corresponding results explored. The findings of the experiments show that the GoogLeNet and AlexNet in multi-label plant disease classification tasks work better.