Explainable AI-Enhanced Deep Learning for Pumpkin Leaf Disease Detection: A Comparative Analysis of CNN Architectures

ResNet50 achieved 90.5% accuracy in identifying five types of pumpkin leaf diseases (downy mildew, powdery mildew, mosaic disease, bacterial leaf spot, and healthy leaves), demonstrating that AI can match or exceed human diagnostic capabilities for agricultural disease detection.

Explainable AI (XAI) techniques were essential for building trust: The researchers used four different visualization methods (Grad-CAM, Grad-CAM++, Score-CAM, and Layer-CAM) to show why the AI made specific decisions, making the system more transparent and trustworthy for farmers and agricultural experts who need to understand and verify the model's recommendations.

The dataset's real-world diversity matters: The 2,000 high-resolution images were collected from multiple agricultural fields rather than controlled laboratory settings, which helps ensure the AI system will work reliably when deployed in actual farming conditions rather than failing on unfamiliar variations.

Multiple AI architectures were tested for fairness: Rather than relying on a single model, the researchers compared seven different deep learning architectures (DenseNet variants, ResNet variants, Xception, and InceptionResNetV2) to identify the best performer, demonstrating responsible AI practice by avoiding bias toward any single approach.

Earlier disease detection could reduce pesticide use and crop loss: By enabling faster, more accurate diagnosis of pumpkin leaf diseases, this AI system could help farmers intervene sooner, potentially reducing unnecessary chemical treatments and preventing widespread crop damage.