Hybrid (VGG16, ResNet50, InceptionV3 and MobileNet) Model for Wheat Leaf Diseases Detection

Abstract

Wheat is one of the most essential staple crops worldwide, and its productivity is often threatened by various diseases affecting its leaves. Early and accurate detection of these diseases is crucial to mitigate crop loss and ensure food security. This study proposes a hybrid deep learning model integrating the strengths of four well-established architectures: VGG16, ResNet50, InceptionV3, and MobileNet, for the efficient detection and classification of wheat leaf diseases.
The hybrid model leverages feature extraction capabilities from each network, combining them into a unified framework to enhance accuracy and robustness. Preprocessed images of wheat leaves from a curated dataset were used to train and validate the model. Features extracted by each architecture were concatenated and processed through a fully connected layer for final classification. The proposed system was benchmarked against standalone models, demonstrating superior performance in terms of accuracy, precision, recall, and F1-score.
By utilizing the complementary strengths of these architectures—VGG16's detailed feature capture, ResNet50's skip connections for deeper learning, InceptionV3's multi-scale analysis, and MobileNet's lightweight efficiency—the hybrid model achieves high detection accuracy while maintaining computational efficiency. This model holds significant promise for practical deployment in agricultural systems, aiding farmers and researchers in real-time disease monitoring and management.