Carbon Robotics, a U.S.-based agritech firm focused on agricultural artificial intelligence and robotics, announced on February 2 the launch of the world’s first Large Plant Model, a groundbreaking AI system for plant detection and identification that ushers laser weeding into a new era of cognitive intelligence in global agriculture.
Trained on the world’s largest, most diverse and fastest-growing agricultural dataset to date, the model features 150 million annotated plants across global crop types, weed species, soil variations, climate conditions and plant growth stages. This extensive training enables farmers to deploy laser weeding on any field and for any crop within minutes, with the model supporting simple and fast personalization based on individual farm and crop conditions, alongside real-time behavioral updates. The innovation makes it easier than ever for farmers to apply cutting-edge AI to field operations and agricultural production.
The Core of Agricultural AI Decision-Making
The Large Plant Model serves as the foundation of Carbon AI, the decision-making brain powering all of Carbon Robotics’ product lineup. Carbon AI processes massive volumes of plant and farm data to make real-time decisions, including weed identification and removal, adapting to diverse field conditions and crop changes. Currently, it provides technical support for the company’s LaserWeeder fleet and Carbon ATK, an autonomous driving kit for existing tractors, empowering farmers with more intelligent and precise weed control as well as reliable autonomous tractor operations.
As the global LaserWeeder fleet operates in fields daily, the system continuously collects real-world data, creating a powerful data flywheel effect. This effect exponentially enhances the model’s capabilities and elevates the performance of LaserWeeder units worldwide, laying an unprecedented foundation for accurate plant identification and data-driven agricultural decision-making.
Paul Mikesell, Founder and CEO of Carbon Robotics, noted that instant plant recognition and real-time behavioral adjustment of the robots allow farmers to maximize the value of the technology on day one. “The Large Plant Model delivers state-of-the-art AI to farmers, optimizing LaserWeeder’s weeding quality for their unique farming environments,” Mikesell said.
Minute-Level Model Personalization
A key advantage of the Large Plant Model is its ability for instant learning of new plant species, a stark contrast to traditional agricultural AI systems that require re-annotation and retraining for new weeds. Previously, identifying and updating the system for a new weed species would take approximately 24 hours; the new model, however, comprehends plant characteristics at a deeper structural level, enabling it to identify and classify even unencountered species accurately.
To complement this capability, Carbon Robotics has launched Plant Profiles, a feature that allows operators to add new plants to the model’s recognition system by selecting just 2 to 3 images on the terminal interface, with real-time optimization of weeding strategies. Already in use at select farms, the feature enables model adjustments in minutes, unlike other AI systems that demand extensive farmer input and take weeks or even months to build new models.
Farm management at Bland Farms in the U.S. confirmed the platform’s significant performance improvement in onion seedling and transplanting scenarios, highlighting its intuitive user interface that allows on-site staff to complete model adjustments quickly and efficiently.
Expanding Global Data Scale
The Large Plant Model is built on photos and data collected by Carbon Robotics’ robots across 15 countries and more than 100 farms. The company stated that its training dataset is growing rapidly with ongoing global device deployment, which is expected to further boost the model’s ability to identify unknown plant species.
Founded in 2018, Carbon Robotics began large-scale delivery of its robotic products in 2022. The company has raised over $185 million in cumulative funding from leading investors including NVentures, Bond and Anthos Capital.
Carbon Robotics announced that the new Large Plant Model will be rolled out to existing customers via a software update. After the upgrade, farmers can directly specify plants to remove or protect on the device interface, achieving more precise field management. The company plans to showcase the technology at upcoming industry events, including the Berlin Fruit Logistica in Germany and the World Ag Expo in the U.S.
Against the backdrop of global agricultural labor shortages and rising input costs, intelligent agricultural solutions that reduce reliance on manual labor and herbicides while boosting crop yields and consistency are gaining significant market attention. The integration of large-scale agricultural data with foundational AI models, as demonstrated by Carbon Robotics, is driving agricultural robotics toward higher levels of automation and intelligence, according to industry observations.
