AutoML Driven Weeds Detection and Elimination
Using image classification tasks to localize weeds for targeted elimination with drones.
Agriculture is under growing pressure to increase yields while lowering costs. However, climate change and other factors are increasing the amount of weeds in organic agriculture. Weeds posses evolutionary advantages over the breeding plants. They take up most of the nutrients and water from the fields reducing the growth of corn, wheat, and soy, for example. Traditional methods of weed control are destructive to a larger area around the weed plants further diminishing the yields.
By using robots and drones equipped with a camera systems, farmers can take images of their land. With AutoML, farmers can easily build an image classifier to tell them which images contain weeds. This classification allows the farmer to target areas affected by weeds with a high degree of accuracy. This allows farmers to reduce the use of pesticides and ultimately, the damage done to the land.
Modulos AutoML is easy to use and fast in generating high performing image classification models. Therefore, farmers can use a combined image data base for training that can be augmented with new images taken at each farm to be sure to capture the lcoal conditions at the time of operation. A similar system can be established for other tasks such as predicting if a field needs to be watered or not.