Cameras that tell a weed from a crop cut herbicide use by half, at millions of acres of scale
John Deere · 2021–ongoing
What they did
John Deere's See & Spray uses boom-mounted cameras and a real-time computer-vision model to distinguish weeds from crop plants as a sprayer passes over a field, firing herbicide only at the individual weed rather than blanket-spraying the whole row.
What happened
Deere reports 2024 customers saved an average 59% on herbicide across corn, soy, and cotton (an independent Iowa State University study measured 76% product savings and $15.70/acre economically). In 2025, adoption scaled to more than 5 million acres, with customers reducing non-residual herbicide use by nearly 50% on average — saving roughly 31 million gallons of herbicide mix for the season.
The so-what
A narrow, well-defined computer-vision task (weed vs. not-weed, in real time, from a known camera angle) with a direct, measurable input-cost payoff is the profile that tends to survive contact with a real growing season, unlike broader "AI for farming" pitches with fuzzier metrics.