A CNN-based Model to Detect and Classify the Tomato Leaf Disease using Transfer Learning

To ensure healthy food production and prevent plants from earlier diseases, efficient and accurate leaf disease detection is essential. Therefore, innovative plant leaf disease detection techniques are urgently needed to mitigate the plant losses caused by illness. The goal of this paper is to identify tomato leaf diseases by classifying healthy and diseased tomato leaves with an enhanced Convolutional Neural Network and a pre-trained CNN model, Inception V3. The proposed Tomato Plant Disease Detection Inception (TPDDI) models are trained by using an open-source database (PlantVillage), which contains a total of 16013 images. The models were investigated with a learning rate of .001, and the experiment showed a better result. In addition, a CNN, VGG16, and InceptionV3 model has been trained and compared with the proposed model. A statistical study highlights the effectiveness of this proposed method, exceeding prior benchmarks. It attains remarkable classification accuracy, precision, recall, and AUC values, resulting in a total accuracy of 99.45%. The elevated performance rate indicates the potential for employing CNN models in the diagnosis of tomato diseases in both field and laboratory settings. This approach can be further integrated into a smart system that could detect diseases earlier and provide an efficient solution to monitor plant health and mitigate crop losses.