The Center for Data Innovation recently spoke with Donát Posta, CEO of Scoutlabs, a Hungary‑based company using a computer vision system and an AI model to monitor insect populations on farms. Posta described how the company’s digital trap system captures daily images, identifies insects with computer vision, and converts the results into structured data that support timely, data‑driven crop management.
David Kertai: What inspired Scoutlabs’ creation?
Donát Posta: Scoutlabs started with a practical problem on my family’s farm. We wanted to stop blanket spraying harsh chemical pesticides and switch to targeted, natural pest control. This means using biological treatments, like bacteria or beneficial insects, that kill specific pests without damaging the surrounding environment.
The problem wasn’t whether these methods work. It was knowing when to use them. These biological treatments are fragile and only stay effective for a day or two after they are applied. They have to be used at exactly the right moment, which is when the pest larvae first hatch or the adults first arrive. Farmers use pheromone traps to detect that moment. These are sticky boxes that lure and catch specific insects to show what is in the field. But checking those traps by hand is too slow. Farmers often miss the narrow 48 hour window when a biological treatment would actually work.
When that happens, the treatment fails. It is not because the biological method is ineffective, but because it was applied too early or too late. This failure is why farmers fall back on heavy chemical sprays that stay toxic for weeks and do not depend on precise timing.
We fixed this by putting a camera and AI into the trap. The system identifies and counts the pests every day then sends the data straight to the farmer’s phone. That turns guesswork into a clear signal. With that daily visibility, farmers can act at the exact right moment. For the first time, they can truly rely on these targeted biological methods instead of defaulting to heavy chemical spraying.
Kertai: How exactly does your digital trap system work in the field?
Posta: Each trap captures a high‑resolution image of the sticky surface every day and sends it to our platform. A computer‑vision system detects the insects in the image, and an AI model identifies the species and converts the results into time‑stamped population data. Species‑level identification matters because different pests require different responses.
Growers receive alerts and simple trend summaries through a dashboard or mobile app. They only need to service the trap when the pheromone lure expires, typically every one to six months, depending on how many insects it captures. This dramatically reduces manual labor and provides a consistent, near real‑time view of pest activity that integrates with existing farm management tools.
Kertai: What kind of data does Scoutlabs collect?
Posta: We collect daily, trap‑level data, including species‑specific insect counts, time‑stamped images, trap location, environmental conditions such as temperature and humidity, and basic device‑health information like whether the sticky surface needs to be replaced. Our goal is to generate reliable biological data at a cost that allows farmers to deploy many traps across a field.
We often have farmers deploy a large number of traps across a field to create a more detailed picture of where insect populations are starting to build up. Farmers place the traps themselves, but they only see emerging hotspots once each trap begins sending daily data. With many traps reporting at once, they can track how populations shift across a field over time—through our online dashboard or mobile app—and respond before those hotspots turn into damaging outbreaks.
Kertai: How do you train your AI models for insect identification?
Posta: When we begin monitoring a new insect species, we focus first on generating reliable data. Our traps capture daily images, and trained entomologists and agronomists review and label each one. Over three to six months, this creates a clean, representative dataset. Once we have enough examples, we train a dedicated computer‑vision model and shift to AI‑assisted identification with full expert oversight. The AI system performs the initial detection, and our team checks the results and corrects any mistakes, creating a continuous learning loop that steadily improves model performance.
After another three to six months, depending on deployment scale, we typically reach 95 to 97 percent identification accuracy. At that point, the AI system handles most classifications, and our team focuses on quality assurance and targeted validation. Importantly, farmers receive dependable data from day one; they never need to wait multiple seasons for useful insights. As accuracy improves, the system scales efficiently, allowing monitoring networks to expand from hundreds to thousands of traps without adding equivalent labor.
Kertai: What is your vision for Scoutlabs?
Posta: Our goal is to make biological crop protection economically viable and environmentally sustainable, giving farmers the confidence to use these tools as reliably as they use chemical pesticides today.
This transition is becoming increasingly urgent as invasive pests spread faster due to global trade and climate change. Our long‑term vision is to support real‑time early‑warning systems that identify new pests quickly and help contain outbreaks before they spread across regions. We see digital pest monitoring as essential infrastructure—providing farmers and policymakers with the timely, actionable intelligence they need to respond to emerging threats before they escalate.
