A deep learning object detection approach for smart pest identification in vineyards

Abstract

Flavescence dorée (FD) poses a significant threat to grapevine health, with the American grapevine leafhopper, Scaphoideus titanus, serving as the primary vector. The disease causes yield losses and high production costs every year due to mandatory insecticide treatments, infected plant uprooting, and replanting. Another potential FD vector is the mosaic leafhopper, Orientus ishidae, commonly found in agroecosystems. The current monitoring approach, which involves periodic human identification of yellow sticky traps, is labor-intensive and time-consuming. Therefore, there is a compelling need to develop an automatic pest detection system leveraging recent advances in computer vision and deep learning techniques. However, the development of such a system has been hindered by the lack of effective datasets for training.

To fill this gap, we created a comprehensive dataset of digitized traps and conducted a detailed labeling process with the assistance of entomologists. Then, we used state-of-the-art architectures, including YOLOv8 and Faster R-CNN, to train object detection models capable of accurately identifying these vectors on yellow sticky traps. Image pre-processing involved automatic cropping to remove irrelevant background information and feature enhancement to improve dataset quality. Our tests evaluated the impact of various factors, such as image resolution, data augmentation and class-specific detection performance.

Results showed the superiority of the YOLO detector in both accuracy and speed, achieving an mAP@0.5 of 92%, an F1-score of approximately 90%, and an mAP@[0.5:0.95] of 66%.

The best-performing model was deployed into the DigiAgriApp platform, an open-source client-server application for centralized farming data management. This application offers farmers a user-friendly tool for real-time vector identification via smartphones while enabling continuous updates to the dataset through citizen science contribution.