Portuguese 
English 
Spanish 
3 min of reading 
0 comments 
Trained With A Database Of 150 Million Plants, The Large-Scale Plant Model Presented By An American Startup Begins To Operate Agricultural Robots Capable Of Identifying Crops And Weeds, Learning Continuously In The Field And Executing Laser Weeding In Different Productive Environments
The American startup Carbon Robotics presented the first Large-Scale Plant Model, trained with 150 million plants, to operate LaserWeeder robots, identify weeds, and perform laser weeding across different crops, with continuous adaptation in the field.
Large-Scale Plant Model Trained With 150 Million Plants
Carbon Robotics claims to have developed the world’s first Large-Scale Plant Model, based on artificial intelligence and trained with a dataset of 150 million labeled plants. The system was created to transform the way agricultural crops are managed on a large scale.
The model feeds the LaserWeeder robots, enabling them to identify and remove weeds using lasers in virtually any crop or field. According to the company, recognition occurs within minutes, regardless of the type of crop or environment.
-
A farm in the Mexican desert has become so green that today it has rivers flowing year-round and even rain in the dry season, all thanks to a technique that transforms dead sand into fertile soil with manure, mountain microorganisms, and what farmers call the secret soil formula.
-
Stable fly infestation frightens producers in Costa Rica, disrupts cattle routine, reduces milk, and puts vinasse at the center of rural controversy.
-
China banned free grazing of goats and sheep in parts of the most eroded plateau on the planet and, along with terraces and thousands of dams, managed to green slopes that were dumping 1.6 billion tons of soil per year into the Yellow River, in one of the largest environmental recoveries in history.
-
China suspended beef purchases from three Brazilian meatpacking plants of JBS, Prima Foods, and Frialto after finding a banned synthetic hormone in tests conducted on the meat, two days after reauthorizing three other plants that had been embargoed since last year.
Continuously learning from data collected by a global fleet of machines, the Large-Scale Plant Model adjusts its performance in real time. This approach allows the system to evolve with each new operation performed in different regions and agricultural conditions.
System Learns From Global Fleet And Shares Improvements
The Large-Scale Plant Model was designed to continuously improve from real-world data.
Each deployment of the LaserWeeder sends new images of plants to the central system, expanding detection and classification capabilities across multiple scenarios.
This feedback loop creates a cumulative data effect, in which performance improvements are shared across the fleet. Thus, advances do not remain restricted to individual machines, as described by the company to Quantum News.
Carbon Robotics emphasizes that the model does not function as a static system. It acts as the core of Carbon AI, a framework that supports both the LaserWeeder and the Autonomous Tractor Kit, integrating real-time operational decisions in the field.
Cost Reduction And Less Dependence On Herbicides
According to the Seattle-based company, using the Large-Scale Plant Model can help farmers reduce labor costs, decrease the application of chemical herbicides, and maintain or increase crop productivity.
The combination of precise identification and laser removal aims to reduce the need for manual weeding and chemical inputs. The company claims that automation allows for operational gains without compromising the quality of agricultural management across different crops.
“When our robots can immediately recognize any plant in any field and adapt their behavior in real time, farmers get the most value from the machines,” said Paul Mikesell, founder and CEO of the company, in a statement.
Plant Profiles Allow Quick Personalization In The Field
In addition to the Large-Scale Plant Model, Carbon Robotics introduced Plant Profiles, a feature that allows for the rapid customization of the LaserWeeder’s operation according to specific crops, weeds, and field conditions.
Available across the entire LaserWeeder line, the tool uses a tablet interface. By selecting two or three representative images via an iPad app, operators can immediately adjust the system’s behavior.
The company states that this real-time personalization reduces the need for prolonged setups or complex reprogramming. Compared to other AI-based agricultural systems, the feature was designed for speed and operational simplicity.
Adjustments That Once Took Months Now Take Minutes
According to Carbon Robotics, adaptations that would typically take weeks or months can now be completed in just a few minutes. This reduces machine downtime and simplifies field operation while maintaining precision in different environments.
By facilitating personalization, the system aims to allow farmers to derive value from autonomous weeding technology more quickly.
The Large-Scale Plant Model remains the foundation for these rapid and continuous adaptations.
“We use plant profiles in seed beds, transplants, and directly sown onions. This has revolutionized our work,” stated a manager from Bland Farms, highlighting the platform’s simplicity and real-time performance in the field.
-
-
3 people reacted to this.