Performance of deep learning vs machine learning in plant leaf disease detection

Abstract

Abstract Plants are recognized as essential as they are the primary source of humanity's energy production since they are having nutritious, medicinal, etc. values. At any time between crop farming, plant diseases can affect the leaf, resulting in enormous crop production damages and economic market value. Therefore, in the farming industry, identification of leaf disease plays a crucial role. It needs, however, enormous labor, greater preparation time, and comprehensive plant pathogen knowledge. For the identification of plant disease detection various machine learning (ML) as well as deep learning (DL) methods are developed & examined by various researchers, and many of the times they also got significant results in both cases. Motivated by those existing works, here in this article we are comparing the performance of ML (Support Vector Machine (SVM), Random Forest (RF), Stochastic Gradient Descent (SGD)) & DL (Inception-v3, VGG-16, VGG-19) in terms of citrus plant disease detection. The disease classification accuracy (CA) we received by experimentation is quite impressive as DL methods perform better than that of ML methods in case of disease detection as follows: RF-76.8% > SGD-86.5% > SVM-87% > VGG-19–87.4% > Inception-v3–89% > VGG-16–89.5%. From the result, we can tell that RF is giving the least CA whereas VGG-16 is giving the best in terms of CA.