Plant disease detection and classification through digital images is an extremely difficult task. Machine Learning (ML) has already been used to detect and diagnose leaf diseases in plants in the last decade. Currently, Convolutional Neural Network (CNN) has demonstrated extraordinary success in the identification and classification of plant diseases. However, choosing a specified CNN architecture with a large number of hyperparameters to get a better result is a very time-consuming process and also it needs expertise in neural networks, which is also considered a challenging task to manually optimize hyperparameters. Hence, Particle Swarm Optimization (PSO) is used to optimize the hyperparameters of CNN and the proposed hybrid PSO-CNN is able to effectively classify the leaf disease. The experimental result of this proposed model is compared with the other three pre-trained models, namely MobileNetV2, VGG16, ResNet50. Simulation results showcased the efficiency of the proposed method over others in identifying and classifying the leaf disease.