Crop Infestation Classification Using MIL-Attention Based CNN

Abstract

Agriculture constantly faces various challenges including attacks from new pests and insects. With large farm sizes and plummeting manpower in the agricultural sector, it becomes challenging to continuously monitor crops for pest infestation. In this research paper, a specific type of pest attack known as the white fly attack has been investigated which affects a variety of crops. This paper presents a multiple instance learning based deep learning approach based on Convolutional Neural Networks for the detection of whitefly pests. A comparative analysis with conventional machine learning and deep learning techniques has also been presented. The performance of the proposed system has been evaluated in terms of the classification accuracy. The experimental results obtained show that the proposed technique attains a classification accuracy of 98.5% and outperforms both separate feature trained machine learning models and well as baseline deep learning models in terms of classification accuracy. Keywords: Precision Agriculture, Whitefly Pest Detection, Feature Extraction, Machine Learning, Deep Learning, Multi Instance Learning (MIL), Classification Accuracy.