Real-Field CNN Detection of the Olive Fruit Fly: Towards Smarter Pest Monitoring in Olive Orchards

The olive fruit fly (Bactrocera oleae Rossi, 1790) is the most damaging pest of olive cultivation, threatening both yield and quality of olive oil across the Mediterranean. Early and accurate detection of adult flies is a cornerstone of integrated pest management, yet conventional monitoring methods based on manual trap inspection remain slow and labor-intensive. In this study, we developed and validated a convolutional neural network (CNN) model trained exclusively on a real-field dataset of 4,278 images collected in olive orchards in Zadar County, Croatia. The model was designed to distinguish B. oleae adults from other insects captured on adhesive traps. It achieved a mean average precision (mAP) of 0.74, with particularly strong performance for the olive fly class (AP = 0.81) and high accuracy at Intersection over Union (IoU) = 50% (AP50 = 0.95). Our results demonstrate that CNN-based detection models trained on field data can provide fast, reliable, and scalable pest monitoring solutions. This approach holds promise for reducing the reliance on manual inspections and supports the development of more sustainable and precise pest management strategies in olive production.