A Novel Deep Learning Design of Plant Disease Recognition and Detection using VGG19, ResNet50, and DenseNet169

The first step in effectively and precisely preventing plant disease in a complex environment is to identify the diseased plants. The identifying of plant diseases becomes digitalized and data-driven only with rapid development of smart farming, enabling better decision support, clever analyses, and planning. In order to increase accuracy, generality, and training effectiveness, a deep learning design of plant disease recognition and detection has been developed using VGG19, ResNet 50, and DenseNet 169. The CLAHE method is first used to identify and localize its leaves in a complex environment. Images that have been segmented using the CLAHE algorithm's findings include the feature of symptoms. Once the leaves have been segmented, they are included in the transfer learning procedure and trained with the data set of sick leaves on a white background. The model is also tested against rust, black rot, and bacterial plaque diseases. Results indicate that DenseNet is more accurate than the conventional method (98.11%), which helps ensure agriculture's long-term viability by reducing the effects of disease on crop yields. Therefore, the deep learning algorithm presented in this paper has important implications for environmental protection, smart farming, and food production