Deep Learning based Bean Leaf Lesion Classification utilizing EfficientNetV2-S

Many areas rely heavily on bean plants, hence treating diseases in these plants calls for early and accurate diagnosis of leaf lesions. Large-scale farming is excluded from traditional approaches of plant disease identification since they are labour-intensive and mostly dependent on professional knowledge. This research study proposes an automated bean leaf lesion detection system based on the EfficientNetV2-S deep learning model. Using modern methods in image processing and machine learning. Our approach classifies three frequent forms of bean leaf lesions: healthy, angular leaf spot, and bean rust. A large collection of labelled bean leaf images is constructed using data augmentation to improve model adaptability. This dataset was used for training and validation of the EfficientNetV2-S model, therefore obtaining a training accuracy and validation accuracy. Important phases of the proposed approach comprised image pre-processing, using cutting-edge data augmentation methods, and hyperparameter optimization to raise performance of models. In terms of accuracy and computing economy, the EfficientNetV2-S model clearly beats conventional approaches and other neural network designs. Furthermore, underlining our model's promise for practical agricultural uses is its great generalization capacity over many environmental circumstances and leaf looks. Emphasising the efficacy of the EfficientNetV2-S model in the automated detection of bean leaf lesions, this work offers a scalable and efficient approach for disease prevention in agriculture. Future research will concentrate on growing the dataset, including real-time detecting features, and investigating how this system might be included into current agricultural monitoring systems.