Potato Leaf Disease Detection by using Deep Learning and Heatmap

Potato is one of the most important food crops worldwide, but its yield is significantly threatened by various leaf diseases, which necessitates proper and timely detection for its effective management. This paper is presents a robust deep learning framework that has been developed for the precise identification of common potato leaf diseases, early blight, and late blight. We are present our method which based on Convolutional Neural Networks renowned for superior performance in automatically extracting hierarchical and discriminative features, and Vision Transformers (ViTs). This is involves in the visual heatmap visualization using Gradient-weighted Class Activation Mapping (Grad-CAM) applied to the CNN path, and Attention Maps derived from the ViT path. This interpretable approach ensures a highly accurate disease classification while providing clear visual explanations, allowing agricultural people to rapidly understand and verify the diagnostic reasoning of the model.