Maize Disease Detection Using CBAM-Enhanced Deep Residual Networks

Maize is a critical global crop, yet its productivity is continually threatened by diseases that reduce both yield and quality. Traditional disease diagnosis, which relies on manual inspections, is subject to human error and inefficiency. In this paper, we present a novel deep learning framework that integrates Convolutional Neural Networks (CNNs) with a Convolutional Block Attention Module (CBAM) and residual blocks to automatically classify maize diseases. The model is trained on the publicly available Plant Village dataset (via Kaggle) and distinguishes among four classes: Blight, Common Rust, Gray Leaf Spot, and Healthy Leaf. By leveraging extensive data augmentation, a cosine annealing learning rate scheduler, and TPUbased distributed training, our approach achieves an overall test accuracy of 97%. Comprehensive experimental results, detailed performance metrics, and confusion matrix analysis demonstrate the superiority of our model, underscoring its potential as a robust diagnostic tool for precision agriculture.