Prediction of Affected Leaf of the Plant using Machine Learning

With the success of deep learning mechanisms, applications that automatically diagnose plant disease have been developed. When used with unknown test datasets, these systems are prone to overfitting, and their diagnostic performance degrades dramatically. From the attention techniques presented in this paper, a new image-to-image identification tool named LeafGAN has been developed. LeafGAN is designed to provide a wide range of healthy image modifications by emerging as a data multiplication tool to improve the efficiency of plant pathogenesis. The proposed method can only change significant portions from images with various backdrops in order to improve the diversity of training images. There are five different types of cucumber illness classification models. In order to reveal that, vanilla CycleGAN is used. Further it is ineffective in improving the data. Research using five types of cucumber disease classification reveals that the data multiplication with vanilla cyclone is ineffective and it is only improving generalization and increasing 0.7% based on disease identification performance. On the other hand, LeafGAN upgrades the diagnostic efficiency by 7.4%, and this research work firmly agrees that the image created by LeafGAN will be of higher quality and reliability when compared to the image created by Vanilla LeafGAN.