Abstract
Pumpkin leaf diseases are significant threats to agricultural productivity,requiring a timely and precise diagnosis for effective management. Traditionalidentification methods are laborious and susceptible to human error,emphasizing the necessity for automated solutions. This study employs on the"Pumpkin Leaf Disease Dataset", that comprises of 2000 high-resolution imagesseparated into five categories. Downy mildew, powdery mildew, mosaic disease,bacterial leaf spot, and healthy leaves. The dataset was rigorously assembledfrom several agricultural fields to ensure a strong representation for modeltraining. We explored many proficient deep learning architectures, includingDenseNet201, DenseNet121, DenseNet169, Xception, ResNet50, ResNet101 andInceptionResNetV2, and observed that ResNet50 performed most effectively, withan accuracy of 90.5% and comparable precision, recall, and F1-Score. We usedExplainable AI (XAI) approaches like Grad-CAM, Grad-CAM++, Score-CAM, andLayer-CAM to provide meaningful representations of model decision-makingprocesses, which improved understanding and trust in automated diseasediagnostics. These findings demonstrate ResNet50's potential to revolutionizepumpkin leaf disease detection, allowing for earlier and more accuratetreatments.