Differential Evolution based compression of CNN for Apple fruit disease classification

Abstract

Apple is one of most favourite fruit all over the world. It may get infected due to various diseases such as blotch, scap and rot. This leads to wastage and loss of apple production which incur financial loss to the farmers. It may also have adverse affect on health of people if they consume infected fruits. Thus a deep learning and machine learning based approaches have been investigated to detect the disease at an early stage for its timely treatment. The best accuracy of 96.87% was obtained using deep learning with proposed convolution neural network (CNN) model haing three convolution layers. The CNN models were also compressed using Differential Evolution (DE)-based process and maximum compression of 82.19% was obtained for VGG16 model without any significant loss in performance. © 2022 IEEE.