Classification of Pear Leaf Diseases Based on Ensemble Convolutional Neural Networks

Over the last few years, impact of climate change has increased rapidly. It is influencing all steps plant production and forcing farmers to adapt their crop management practices using new technologies based on data analytics. This study aims classify diseases images collected directly in field deep learning. To this end, an ensemble learning paradigm investigated build a robust network order predict four different pear leaf diseases. Several convolutional neural architectures, named EfficientNetB0, InceptionV3, MobileNetV2 VGG19, were compared ensembled improve predictive performance by adopting bagging strategy weighted averaging. Quantitative experiments conducted evaluate model DiaMOS Plant dataset, self-collected dataset field. Data augmentation was adopted generalization model. The results, evaluated with range metrics, including accuracy, recall, precison f1-score, showed that proposed outperformed single classifying real field-condition variation brightness, disease similarity, complex background, multiple leaves.