The health of seed potato crops is being threatened by viruses like the Potato Virus Y (PVY). In the last five years, more than 15% of harvested seed potatoes and sugar beans were infected with the Potato Virus Y [1]. There is an empirical relationship between the number of aphids and the number of infected plants [1]. Early detection of aphids creates the opportunity to instantly use pesticides. Currently, manually counting aphids is very time-consuming and the use of object detection models, to detect aphids and accordingly decide the necessity of applying pesticides, could offer a solution. This study considers a single-stage object detection model for the real-time detection of aphids on yellow sticky plates with YOLOv5. A dataset has been acquired with images of yellow sticky plates with aphids and other insects. High-resolution images are being utilized to capture the small size of aphids. To minimize computing power, a tiling technique is being explored. This involves dividing the images into smaller sections called tiles. The network is competent in selecting tiles that always contain either an aphid or an insect, named positive tiling. The YOLOv5l-model with positive tiling is the best-performing model with an F1-score of 0.525. Possible extensions of this study are being discussed, together with suggestions for future research.