Hybrid CNN–Transformer Model for Maize Leaf Blight Classification Using Adaptive Genetic Optimization

Abstract

Maize leaf blight is a disastrous foliar disease in the world production of maize that causes significant losses in terms of yield annually. The classical machine learning (ML) and convolutional neural network (CNN) models often exhibit poor generalization under diverse field conditions due to variations in illumination, background complexity, and leaf morphology. To address these challenges, this study proposes a hybrid CNN–Transformer architecture optimized using Adaptive Genetic Optimization (AGO) for accurate maize leaf disease classification. The hybrid model utilizes CNN-based spatial feature extraction and global self-attention mechanism of the Vision Transformer (ViT) to capture both local and contextual patterns of diseases. The AGO algorithm dynamically optimizes key hyperparameters, including learning rate, batch size, embedding dimension, and attention heads, according to the population diversity and fitness evaluation, thereby improving convergence speed and classification performance. Experimental analysis conducted on an augmented maize leaf disease dataset demonstrated that the proposed model achieved an overall classification accuracy of 95.7%, outperforming conventional architectures including VGG16, ResNet50, DenseNet201, MobileNetV3, and a baseline ViT. Ablation studies, statistical stability analysis, and robustness evaluation further confirmed the effectiveness, reliability, and generalization capability of the proposed framework under varying field conditions. The proposed AGO-CNN–Transformer framework provides an effective and computationally feasible solution for intelligent maize disease diagnosis and precision agriculture applications.