Carbon Robotics’ Large Plant Model Brings Real-Time Plant ID to Laser Weeding

Carbon Robotics, the Seattle startup behind the LaserWeeder – a robot fleet that uses lasers to eliminate weeds – has introduced a new AI system called the Large Plant Model (“LPM”). The company says LPM can detect and identify plant species in real time, helping the LaserWeeder make more accurate decisions about what to remove and what to leave behind. The company has raised more than US$60 million from investors including Sozo Ventures, Fuse Venture Capital and Anthos Capital.

Weeds are moving targets. What count as weeds vary by crop, soil and season. Even the same weed can look different depending on growth stage and field conditions. Paul Mikesell, the CEO of the company, says the traditional workflow required creating new labels and retraining models whenever a new scenario appeared – often taking roughly 24 hours per update.

With LPM, farmers can teach the system faster. The model was trained on 150 million labeled plant photos and data points collected across more than 100 farms in 15 countries. The new model is delivered via software update, and farmers can use the robot’s interface to review images captured in the field and specify what to kill versus what to protect without waiting for a retraining cycle.