CucuNetCNNs: Application of novel ensemble deep neural networks for classification of cucumber leaf disease

Abstract

The accurate diagnosis of plant diseases is crucial for improving agricultural productivity and ensuring global food security. This study introduces an advanced approach to cucumber leaf disease classification by integrating novel deep learning methodologies. Two custom-designed convolutional neural networks (CucuNet-CNN1 and CucuNet-CNN2) are proposed, alongside pre-trained models such as InceptionResNetV2, EfficientNetV2M, and NASNetMobile, to classify various disease types. To enhance classification performance, an ensemble model (5-EnsCNNs) is developed, combining the strengths of these architectures. Additionally, a Spiking Neural Network (SNN), inspired by neuromorphic computing principles, is employed. Experimental results show that the SNN achieves a remarkable accuracy of 98.91 % in classifying six cucumber leaf diseases, surpassing the performance of individual and ensemble models. The integration of novel CNN architectures, ensemble strategies, and SNN-based methods represents a significant advancement in automated plant disease diagnosis, paving the way for more accurate and reliable agricultural diagnostics.