How do farmers customize the weeder?

Farmers use an iPad app to review thumbnails from their fields and tag a small number of plants as 'crop' or 'weed.' The model adjusts immediately without needing to retrain or download new software.

Carbon ATK is an autonomy kit that retrofits existing John Deere 6R, 8R, 8RX, 8RT (2019+) series tractors with cameras and sensors for obstacle avoidance, path planning, and fully autonomous weeding missions.

Why did Carbon Robotics partner with iMerit?

Carbon Robotics partnered with iMerit to create training data. iMerit labeled a million plant images, while Carbon did about 2,000 on its own, helping to build the large plant model more efficiently.

Large Language ModelsRoboticsThe Robot Report2h agoAug 28, 2026, 01:00 JST2 min read

Carbon Robotics launches single large plant model for instant weeding customization

Send on LINE

Key takeaway

Carbon Robotics introduced a large plant model that lets farmers customize its laser weeder instantly by tagging examples.
The model is pretrained on millions of images and adjusts behavior without new software downloads.
Farmers can switch between crops with minutes of setup.

3 Key Points

What happened
Carbon Robotics replaced its crop-specific AI vision models with a single "large plant model" trained on millions of images, letting farmers customize laser weeding via an iPad app by tagging a few plants as crop or weed.
Why it matters
This removes the need for time-consuming retraining when a new crop, weed, or geography appears, so the same global model can zap weeds in a carrot field in Arizona, then switch to lettuce or herbs on another farm with only minutes of configuration.
What to watch
The model was built with data-annotation partner iMerit, which labeled a million images (Carbon did about 2,000 itself). The company also released Carbon ATK, an autonomy kit that retrofits John Deere 6R, 8R, 8RX, 8RT (2019+) tractors for fully autonomous weeding missions.

Ask the AI about this article →

Summaries like this, in your inbox every morning.

Get news like this every morning — free

Context & Analysis

The shift from crop-specific models to a single foundation model marks a practical advance for agricultural AI. Rather than retraining for each new field or crop, the model learns from examples on the fly, which could reduce downtime during the growing season. Carbon Robotics built this capability with iMerit, a data annotation partner that labeled a million images—a task Carbon itself only managed about 2,000 images for before scaling. The iPad-based tagging interface lowers the technical barrier for farmers, who can now define weeds and crops without specialized ML knowledge.

Beyond weeding, Carbon Robotics is expanding into tractor automation with Carbon ATK, a retrofit kit for John Deere tractors. This suggests a broader ambition to automate multiple farming workflows, such as tillage, spraying, and harvesting, using the same autonomy platform. The implications for labor and precision agriculture could be significant, though the article does not provide performance data or adoption numbers. Overall, the news highlights a trend toward flexible, on-device AI that adapts to user feedback, which may become more common across industries.

FAQ

How do farmers customize the weeder?: Farmers use an iPad app to review thumbnails from their fields and tag a small number of plants as 'crop' or 'weed.' The model adjusts immediately without needing to retrain or download new software.
What is Carbon ATK?: Carbon ATK is an autonomy kit that retrofits existing John Deere 6R, 8R, 8RX, 8RT (2019+) series tractors with cameras and sensors for obstacle avoidance, path planning, and fully autonomous weeding missions.
Why did Carbon Robotics partner with iMerit?: Carbon Robotics partnered with iMerit to create training data. iMerit labeled a million plant images, while Carbon did about 2,000 on its own, helping to build the large plant model more efficiently.

The Robot ReportRead Original Article