Autonomous Weed-Control Robots gain real-time plant-recognition AI

Carbon Robotics introduced the Large Plant Model (LPM), an AI trained on more than 150 million labeled plant images to identify crops and weeds in real time. The model was deployed as a software update to existing LaserWeeder robots and lets operators choose which plants to preserve or remove via the onboard interface.

Carbon Robotics has added a new plant-recognition system called the Large Plant Model (LPM) to its LaserWeeder robots. The model identifies plant species and structural traits in real time so machines can distinguish crops from unwanted plants. The update was issued as software to existing LaserWeeder units, which continue to gather field imagery during operation. Farmers use an onboard interface to mark plants to preserve or eliminate based on photos captured by the robots. Key details: - LPM was trained on more than 150 million labeled plant images from over 100 farms in 15 countries. - The model analyzes both species identity and structural traits to inform on-robot decisions. - The software update was rolled out to existing LaserWeeder units, which keep collecting new field images. - Operators interact with the system via an onboard interface to select plants to preserve or target. - The article describes the change as supporting chemical-free fields, reducing downtime for new targets, and lowering the need for manual data labeling. Summary: The LPM gives LaserWeeder robots species-level recognition that can be used to target weeds without pausing operations, according to the article. Undetermined at this time.