Tomato leaf disease detection and classification using CNN with Grad-CAM

Tomato leaf diseases, such as early and late blight, are the primary causes of crop output and quality loss, which emphasizes the importance of promptly and accurately identifying them. To automatically classify leaf images of the tomato into three groups i.e. healthy, early blight, and late blight a Convolutional Neural Network (CNN) model is introduced. In order to perform an appropriate performance evaluation, the dataset was split into three subsets for testing, validation, and training. The created model&s;s overall accuracy was 92%. By displaying the regions that the model considered, these visual explanation tools Gradient-weighted Class Activation Mapping (Gradient-CAM) and Class Activation Mapping (CAM) were put into place to aid the user in understanding the model. These visualizations make it possible to see that the model is not only focusing on the areas of the leaves that are infected but also doing so in the correct way, i.e., thus proving the model&s;s reliability. The experiment&s;s findings demonstrate that the CNN-based approach that is, early disease detection is possible, and as such, it may be regarded as a precision agriculture decision-support tool.