Chinese cabbage leaf disease prediction and classification using Naive Bayes VGG-19 convolution deep neural network

In this research, we delve into the efficacy of employing transfer learning in conjunction with convolution deep neural networks based on the inception architecture for classifying diseases in Chinese cabbage leaves. To combat over-fitting, we employ data augmentation techniques including image scaling, rotation, and translation. We leverage the per-trained parameters trained on extensive datasets and fine-tune it for our specific task. The initial weights of VGG-19 are kept fixed, and a Naive-Bayes component is integrated into the underlying network, resulting in the model VGG-19-Navie-bayes. Relative to advanced frameworks like ResNet-101, Inception-ResNet-V2, Xception, and Darknet-53, our suggested method showcases encouraging outcomes. Notably, the images are not segmented, and our technique attains an average accuracy of 96.63 percent even under more intricate conditions. The study encompasses five distinct types of Chinese cabbage leaves: mildew, club-root, healthy, leaf miner, and back moth. The experimental outcomes affirm that our suggested approach effectively distinguishes between various plant diseases.