The potato virus Y can infect potatoes and heavily impact the strong position Dutch potato farmers have on the market.
Because this virus gets carried by aphids, the usage of pesticides is crucial in order to prevent the food waste of hundreds of potatoes.
Overuse of pesticides can strongly impact the environment. Which is why they can only be used when it is proven that aphids are
present. Determining this will prove difficult: the insects are relatively small and can be mistaken for fruit flies. Computer vision may
offer a solution here. Within this research, two detector models, YOLOv8 and RetinaNet are trained on a dataset containing winged
aphids, wingless aphids and 2 kinds of fruit flies in order to determine which one is the best in detecting aphids in an early state as
possible. Not only will these detectors be trained to be able to detect aphids from other insects, but also to determine exactly what
the currently viewed insect is. The latter proves to be more difficult for the models with YOLOv8 giving a best F1 score of 0.84 and
RetinaNet 0.53. As opposed to just determining an aphid from a non-aphid with F1 scores of 0.90 and 0.76. With this data, it can be
determined that YOLOv8 is the better model for detecting aphids.