Optimizing EfficientNetv2 Model with RandAugment Data Augmentation for Detecting Wheat Diseases in Smart Farming

Abstract

Wheat diseases threaten global food security, necessitating improved detection methods. In this paper, we integrate EfficientNetv2 model and RandAugment data augmentation to accurately and efficiently identify wheat diseases. EfficientNetv2, known for its optimal mix of accuracy and computing efficiency, is reinforced by RandAugment, a versatile data augmentation approach that randomly modifies training data. This augmentation method greatly enhances the model’s generalisation and performance on new data. Our extensive experimentation reveals that this integrated technique improves model accuracy and robustness relative to baseline models. Proposed model gained the 96.73% accuracy on prescribed dataset. The results show that EfficientNetv2 and RandAugment can detect wheat illnesses on a large scale. This could change precision agriculture by enabling early and accurate disease management.