PERFORMANCE EVALUATION OF CORN LEAVES DISEASES CLASSIFICATION USING LIGHTWEIGHT DEEP LEARNING

Abstract

Based on this research we proposed the suitable preprocessing method and changing activation layer in ShuffleNet model for corn leaves disease classsfication. This start from preprocessing images which is enhance contrast and denoise image to show the better method to be use in deep learning models. Then we also changing ReLU layer into ClippedReLU layer which works better in MobileNet-V2 to improve the better performance on ShuffleNet. The dataset used in this research is from Plant Village Dataset in Kaggle database. The result showed that contrast enhance could improve the performance of deep learning models compared to denoise and original images. Also, the ShuffleNet + ClippedReLU layer improve some of training parameters including time but ni the accuracy. This changing activation layer on ShuffleNet needs more exploration to improve the accuracy.