Reducing pesticide use with AI-driven weed control

Robovision and ILVO developed a weed control system that combines vision AI with drones to allow hyper localised spraying.

Using AI to reduce pesticide use up to 80%

Robovision provides agro-robotics with the intelligence needed for in-field weed detection and precision farming. The goal: to see and read any field, any time, no matter the conditions. Next, our tool tells farmers exactly how to optimally treat their crops. We go beyond mere analysis but include automated action by AI (robots).

Figure 1: the Robovision Platform allows plant experts, without a technical background, to label and train an AI model for weed control 

The challenge

A growing global population asks for a growing and increasingly efficient food supply. Sadly the side-effect is excessive use of crop protection chemicals, which is not in line with the necessary shift towards sustainable agriculture. With the deployment of Europe’s New Green Deal, efforts will be accelerated to decrease any unnecessary impact agriculture has on the environment.

Figure 2: Ruben Van De Vijver, Researcher in technology and food at ILVO talked about the ambitious goal of Europe to reduce chemical use in agriculture to 50% by 2050

The solution

Combining several of our AI model’s strengths, Robovision and its partners have built an end-to-end spraying solution. We can now read the field, identify weeds, provide the machinery with a detailed map and allow it to spray without losing or damaging crops. Safe to say the technology is sustainable by design, for creating AI-powered machines radically reduces the use of crop protection chemicals. In other words: good news for both the farmer and the planet.

AI applications in agriculture

Figure 3: the end-to-end spraying solution provides machinery with a detailed map that reduces pesticide use to 80%

The approach

Identifying a broad array of weeds in all possible weather conditions and varying light is challenging, to say the least. More so when you attempt to keep the time between scanning, reading, interpreting and sending the tasks to the machines as short as possible. A carefully chosen training dataset of various fields with various types of weeds and soil and tailored data augmentation allow the model to work with a high level of variability. All day, rain or shine, the identification is close to perfect.

Figure 4: combining drones with an AI-based system to deliver vision intelligence

Reaching Desirable Outcomes


The unique overview and precise identification of weeds allows for a high level of accuracy.


The level of precision, in turn, guarantees a healthy field and maximum crop protection under all weather conditions.


Efficiency is key in sustainability. Our technology supports farmers’ efforts in reducing the use of harmful substances. Better for their wallet and better for the planet.


The model can easily be adjusted, applied for innumerable machines without losing anything of its speed.


The drone flies autonomously, sends the data automatically, which is analyzed instantaneously by our AI model. The task map is then generated and sent to the spraying machine, without any supervision needed.

Ready to power your systems with vision AI?

Talk to an Expert