Utilizing Machine Learning to Detect Whiteflies on Tomatoes in a Controlled Environment

Abstract

The tomato is the vegetable crop with the most economic impact globally, and its production has expanded significantly over time. Recent years have seen the discovery of whiteflies as a serious loss-maker in producing fresh-market and greenhouse tomatoes. Tomato plants are harmed both directly and indirectly by whitefly nymphs and adults. The leaves are attacked by whiteflies, causing them to turn yellow and curl up, resulting in their destruction. Currently, early whitefly migrations are detected using yellow sticky traps. However, executing this activity takes a lot of time and effort. In order to identify plant pests more rapidly and precisely, a method of early detection that significantly reduces economic losses was created. In this research, we proposed to use an image analysis and machine learning technique to develop a model for detecting whiteflies on tomatoes in a greenhouse. Images of leaves covered in whiteflies were obtained, and the EfficientDet-D0 model was used to train the machine learning algorithm. Results indicate that this new method might detect whiteflies with acceptable precision and an F1 score of 0.40, indicating that EfficientDet-D0 models reliably recognize the distinctive characteristics of whiteflies.