Tools
Tools
Shetty, Aniket Kutty (2024) Corn Leaf Disease Detection Using Deep Learning and Explainable AI. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (10MB) | Preview |
Preview |
PDF (Configuration Manual)
Download (3MB) | Preview |
Abstract
Plant health is crucial for maintaining agricultural productivity & food security and automated plant disease diagnostics help us achieving this while benefiting the economy. This research work uses advanced deep learning models to analyse early and accurate detection of Maize leaf diseases which include Blight, Common Rust, and Gray Leaf Spot. In this work, ResNet-50 and MobileNetV2 architectures are employed, along with complex data augmentation and transfer learning to enhance classification capability. Namely Grad CAM++ (Gradient weighted Class Activation Mapping++) and LIME (Interpretable Model Agnostic Explanations), Explainable AI (XAI) methods enhance model interpretability providing graphical visualization of predictions to get away from the black-box character of these models. After Hyperparameter Tuning MobileNetV2 being a light model gave a decent accuracy of 90.16%, to optimize computational power, for restricted environments. While Resnet-50 model provided a wonderful performance with accuracy of 96%. The complementary application of Grad-CAM++ and LIME demonstrates the models’ potential of identifying disease-relevant traits indicating strong confidence and applicability among the agricultural sectors. In this work, they contribute to bridging the gap between deploying conceptually efficient Deep Learning AI models and their functional deployment to promote precision agriculture.
Actions (login required)
 |
View Item |