John Deere, targeting only weeds with AI. A move to change the future of agriculture.

The Challenge of Major Agricultural Machinery Manufacturer John Deere

John Deere is a globally renowned manufacturer of agricultural machinery such as tractors. [4] The company is working to realize "precision agriculture" using AI technology in response to challenges such as increasing global food demand and a shortage of agricultural workers. [11, 19] This article introduces "See & Spray," a core technology in this effort.

The Problem to Be Solved: Herbicide Costs and Environmental Impact

In traditional agriculture, "broadcast spraying" of herbicides over the entire field was common. [4] However, this method not only incurs high costs by spraying chemicals even in areas without weeds but also poses a significant burden on the soil and surrounding environment. [3, 4] Furthermore, the emergence of herbicide-resistant weeds had become a problem. [12] Farmers were seeking new methods to effectively manage weeds, reduce costs, and achieve sustainable agriculture.

How AI Was Used: Real-time Weed Detection System "See & Spray"

John Deere acquired the AI startup Blue River Technology and developed "See & Spray" by advancing its technology. [4] This system uses 36 cameras mounted on the sprayer and powerful NVIDIA processors (GPUs). [1, 17, 21] While moving at approximately 19 km/h, the cameras scan over 2,500 square feet of field per second, and AI-powered computer vision (image recognition) technology identifies crops and weeds in real time. [2, 4] Then, only what the deep learning model determines to be a weed is precisely sprayed with herbicide from individual nozzles. [3, 12]

Implementation Effects and Key Takeaways

• ▸Significant reduction in herbicide use. One study reported an average reduction of approximately 50% in the use of non-residual herbicides. [2, 7] Separately, there are reports of up to a 70% reduction. [26]
• ▸Achieves both cost reduction and increased yield. One trial showed an average reduction of $7.46 per acre in herbicide costs and an average increase of 2 bushels/acre in soybean yield. [8, 9]
• ▸Contribution to sustainability. Reducing the use of chemicals leads to a decrease in environmental impact and protects soil health. [1, 16]
• ▸Utilization of Edge AI. It is crucial that the AI model runs on computers mounted on the vehicle, enabling real-time processing even in farms with unstable internet connections. [1, 17]
• ▸Vast image data improves AI accuracy. The AI is trained with an image library of over 1 million images to enhance its accuracy. [12]

What Japanese Companies Can Learn

John Deere's case demonstrates that significant value can be created by combining deep knowledge of a specific industry (agriculture) with general-purpose AI technology. The idea of using AI to "precisely" solve "common" problems (e.g., broadcast spraying) within one's area of expertise is applicable in various fields, such as defect detection in manufacturing or infrastructure maintenance and inspection. Furthermore, the importance of on-site (edge) data processing for real-time performance and stable operation is also a valuable takeaway.

Today's AI News

Reference Links