John Deere

Conventional broadcast spraying applies herbicide across an entire field indiscriminately, wasting chemicals on bare soil and crops that do not need treatment. This inflates input costs for farmers, drives excess chemical runoff into ecosystems, and accelerates herbicide-resistant weeds. Distinguishing a weed from a crop plant at 12 miles per hour across a 120-foot boom, in variable lighting and dust, is a computer-vision problem that had never been solved at agricultural production scale.

John Deere developed See & Spray, an embedded computer-vision system that uses banks of cameras and onboard GPUs running deep-learning models to distinguish weeds from crops in real time as the sprayer moves through the field. Each of dozens of nozzles is individually actuated within milliseconds to spray only the weeds it detects. The technology, accelerated by Deere's 2017 acquisition of AI startup Blue River Technology, processes imagery on the machine itself — edge AI — because cloud latency is impossible at field speed. See & Spray Ultimate was released commercially for row crops such as corn, soybeans, and cotton.

Total timeline: 2017–2022: from the Blue River Technology acquisition to commercial See & Spray Ultimate deployment on production sprayers

See & Spray demonstrated herbicide reductions of more than two-thirds on non-residual herbicides in field trials — averaging around 60%+ savings — cutting one of a farmer's largest variable costs while sharply reducing chemical load on the environment. It became a cornerstone of John Deere's transformation from an equipment manufacturer into a precision-agriculture technology company, with AI and autonomy positioned as central to future revenue. The system processes millions of plant classifications per acre in real time at commercial field speeds.