Extraction of Heterogeneous Features from Various Fruits and Classification of diseases using Deep Learning Techniques

Abstract

The success of the farming business and the economy as a whole is directly tied to each individual farmer's output. Micro-level illnesses that emerge during fruit development may have a significant influence on output. There is an overwhelming variety of viruses, necessitating improvements in image processing and other informational resources before effective treatments can be proposed. Unparalleled approaches to segmenting, extracting characteristics, and classifying in order to identify overt conditions constitute a substantial portion of the research in this domain. However, in the event of larger and more in-depth datasets, machine learning and artificial intelligence technologies like deep learning and the like are necessary to discover illnesses that would otherwise be invisible to the human eye. Machine designers are not always sure what kinds of illnesses can be identified by mixing algorithms since there are so many possible combinations. In this paper we proposed a fruit disease detection and classification using deep convolutional neural network. The VGG-16 and VGG-19 both deep learning frameworks are utilized for detection of heterogeneous fruits. The both deep learning frameworks are archives 96.10% and 98.10% average accuracy on heterogeneous fruit dataset. In overall analysis the VGG-19 obtains higher accuracy with 200 epoch size which is higher the VGG-16 and other conventional machine learning and deep learning frameworks.