Deep Learning based Pest Classification in Soybean crop using Residual Network-50

Abstract

Agriculture is the fountainhead of human sustenance, farmers have an imminent risk of the crops getting attacked by the pest. Times before the advancement in science and technology, farmers incorporated traditional techniques in dealing with pests, but the major issue faced by them was the detection and classification of the various species of pests. With the advancement of technology, researchers implemented a Deep Learning method to classify various species of pests by analysing pictures captured in real-life situations. In this paper, deep convolutional neural networks (DCNN) are used to classify four different categories of bugs/pests found on the soya bean crops. As Deep Learning outperforms when working with a large data set, various augmentation techniques were applied to the raw images to make a larger dataset and improve accuracy. The results say that the deep learning architecture when fine-tuned can give higher classification accuracy against other traditional classification methods, reaching accuracies up to 96.25%. The results show that the architectures help to understand pest control management in soya bean crop fields.