In the last decade, plant leaf disease identification has been an efficient research subject. In connection with this interest, deep learning architectures show a remarkable era in various fields of image processing and computer vision, including image classification, function detection, and image pattern recognition. In this study, we examine many aspects of convolutional neural networks for image pattern recognition. We examine the early and late fusion of multiple pattern recognition classifiers using various plant leaves. Commonly, it considers disease discovery with the diagnostic technologies available. In standard cases, planters usually do not discover the disease. Therefore, plant leaf disease detection is a significant research problem, and one of their goals is to uncover an effective way to identify leaf image disease. The article has made a potential effort to find a process that should be able to expose plant leaf disease using early and late fusion of two classifiers: modified Optimized Deep Neural Network (ODNN) with different parameters of evolutionary optimization of Grasshopper algorithm (GOA), Speeded Up Robust Features (SURF) and Convolutional Neural Network (CNN) that could support the system to achieve excellent performance. Classification quality parameters are determined, and research to explain the validation of the model has been carried out.