Development of a system for site-specific control of the Colorado potato beetle The orange-red larvae of the Colorado potato beetle are extremely voracious and can destroy a significant part of the leaf apparatus of the potato crop in no time. This can significantly reduce the potato yield. Appropriate, timely control of these larvae is therefore appropriate. However, because these larvae occur in patches in plots, site-specific control can significantly reduce the amount of crop protection required. Therefore, a system for detecting and site-specific treatment of Colorado potato beetle larvae will be developed. For this purpose, we are using object detection and classification through deep learning (artificial intelligence). The larvae (and adult beetles) are spotted on the basis of normal colour photographs. The detection system is mounted on a drone platform to allow accurate detection of Colorado beetle larvae in the potato crop. The specific methodology used in this use case was a YOLOv8 model trained on RGB images from a Sony ILX sensor on a DJI M350 UAV at 10 m altitude. These conditions allowed for high resolution images and feasible detection of both larvae and adult Colorado beetle in a potato crop. Datasets of infested fields will be collected throughout the 2026 growing season at experimental fields of EV-ILVO (if and when beetle/larvae are present). The results and datasets will be published and be made available upon request. The impact we foresee is mainly a reduction in the total amount of crop protection products used for control of the Colorado potato beetles. Based on preliminary results, the models offer a high accuracy but still have to be made more robust for variations in lighting conditions, crop growing stage and larval growing stage. We would be interested in applying the methodology for other pests and crops and would be also interested in Colorado potato beetle datasets to validate our models further.

Source: Annual AgData Use Case Summit 2026 booklet.