Early identification of Tuta absoluta in tomato plants using deep learning

Abstract

The agricultural sector is highly challenged by plant pests and diseases. A high–yielding crop, such as tomato with high economic returns, can greatly increase the income of smallholder farmers income when its health is maintained. This work introduces an approach to strengthen phytosanitary capacity and systems to help solve tomato plant pest Tuta absoluta devastation at early tomato growth stages. We present a deep learning approach to identify tomato leaf miner pest (Tuta absoluta) invasion. The Convolutional Neural Network architectures (VGG16, VGG19, and ResNet50) were used in training classifiers on tomato image dataset captured from the field containing healthy and infested tomato leaves. We evaluated performance of each classifier by considering accuracy of classifying the tomato canopy into correct category. Experimental results show that VGG16 attained the highest accuracy of 91.9% in classifying tomato plant leaves into correct categories. Our model may be used to establish methods for early detection of Tuta absoluta pest invasion at early tomato growth stages, hence assisting farmers overcome yield losses.