ResNet50 Integrated Vision Transformer for Enhanced Plant Disease Classification

This study introduces an innovative hybrid model that integrates the powerful feature extraction capabilities of ResNet 50 with the advanced pattern recognition and contextual understanding afforded by the Vision Transformer (ViT) for the classification of plant diseases. By embedding a class token into the input feature sequence and leveraging a novel multi-head attention mechanism, this model captures both local and global image features, leading to a significant enhancement in diagnostic accuracy and efficiency. Our approach involves a meticulous configuration process that adjusts hyperparameters such as the number of transformer encoder layers and attention heads to optimize performance. Rigorous testing on a meticulously curated and pre-augmented dataset underscores the model’s superiority over existing methodologies, demonstrating its readiness for real-world agricultural applications. The findings from this research not only mark a substantial advancement in plant disease diagnostics but also set a new benchmark for the application of hybrid deep learning models in precision agriculture, paving the way for sustainable farming practices in the face of escalating global food demands.