Advancing Farming Frontiers: How AI in Agriculture Images is Revolutionizing the Field

Researchers at North Carolina State University’s Plant Sciences Initiative (N.C. PSI) are pioneering the use of artificial intelligence to enhance agricultural yields, efficiency, and sustainability. Central to this effort is the application of Ai In Agriculture Images, powered by new robotic tools and high-performance computing, aiming to provide producers with unprecedented insights and capabilities.

June 13, 2024 | 8-min read

Researchers with NC State's BenchBot 3.0 robot used for collecting ai in agriculture images for a large open-source repository.

The Foundation: Building a Massive Agricultural Image Repository

At the heart of N.C. PSI’s initiative lies the creation of what aims to be the world’s largest open-source agricultural image repository. This ambitious project is driven by BenchBot 3.0, a sophisticated robot deployed on NC State’s Centennial Campus. Its mission is to capture hundreds of thousands of plant photos, starting with 500 pots containing various plant species. These images, combined with data from earlier BenchBot versions, form the bedrock for training advanced AI models.

Chris Reberg-Horton, N.C. PSI Platform Director for Resilient Agriculture, emphasizes the critical need for such a resource. He draws a parallel to the development of self-driving cars, which required millions of labeled images to train AI systems to recognize pedestrians, cyclists, and traffic signals. “In a lot of economic sectors, we have literally millions of images labeled… We don’t have that in agriculture,” Reberg-Horton explains. This repository, accessible to technology developers large and small, seeks to fill that gap, enabling the creation of robust Ai In Agriculture Images applications.

The team has already imaged around 30 common North American weed species, along with key cash crops like corn, soybeans, cotton, wheat, rye, barley, and oats using previous BenchBots. BenchBot 3.0 significantly scales up this effort. “In theory we can do about 1,000 plant pots per day,” Reberg-Horton notes, adding that the focus this summer is on weeds, followed by winter cover crops. To diversify the dataset, companion BenchBots are operating at the USDA’s Agricultural Research Center in Maryland and soon at Texas A&M University.

Hardware engineer Mark Funderburk beside an earlier BenchBot prototype used for capturing plant images for agricultural AI training.

Unleashing Potential with Supercomputing Power

Processing and analyzing the colossal dataset generated by the BenchBots requires immense computational power. N.C. PSI has acquired the NVIDIA Grace Hopper 200, the most powerful supercomputer of its kind on NC State’s campus. Capable of handling five terabytes of data per second, this machine is crucial for training complex AI models based on the collected agricultural images.

Jevon Smith, N.C. PSI’s research computing manager, highlights the supercomputer’s significance: “The Grace Hopper… gives us the capability to train even larger-scale models to become more accurate and more predictive in nature and come up with solutions to more complicated challenges.” This computational muscle allows researchers to move beyond simple image recognition towards creating sophisticated predictive tools for farmers.

Jevon Smith with the powerful NVIDIA Grace Hopper supercomputer processing vast datasets of ai in agriculture images at NC State's PSI.

Transforming Pixels into Precision Agriculture Tools

The ultimate goal of leveraging ai in agriculture images is to fully realize the promise of precision agriculture. This approach involves delivering precise inputs—water, fertilizer, pesticides—exactly when and where needed, minimizing waste and environmental impact while maximizing yield. While the concept has existed for decades, the lack of detailed field knowledge has been a major hurdle.

Reberg-Horton points out that while smart equipment capable of variable rate application is available, “we have been stuck on creating enough intelligence to tell that equipment what to do.” He asserts that computer vision, powered by AI trained on vast image datasets, is the key technology to bridge this gap. “Computer vision is the technology that can do it, and we will start seeing cameras on all of our agricultural machinery.”

Smith adds that despite controversy surrounding AI, its application in agriculture holds immense positive potential. “Applications in the agriculture sector include increasing yields, reducing waste, helping reduce carbon emissions, and more. It can be used for great good,” he states. Existing applications already include self-driving tractors, variable rate applicators, water management systems, and tools for optimizing planting and harvest timing.

Diverse AI Applications Taking Root at N.C. PSI

The focus on ai in agriculture images is part of a broader push for AI-driven solutions at N.C. PSI. Other notable projects include:

• Soybean Management: Rachel Vann is leading the development of an AI-powered web app to help growers optimize soybean planting and management across North Carolina.
• Sweetpotato Sorting: Cranos Williams is working with the Sweet-APPS team to use sensors and machine learning for automated post-harvest sorting and grading of sweetpotatoes, reducing labor needs.
• Climate Resilience: Ross Sozzani is collaborating with USDA scientist Anna Locke and researchers in Belgium, using machine learning to identify temperature-tolerant soybean varieties more efficiently, crucial for adapting to climate change.

Practical Tools Emerging from Image Analysis

Reberg-Horton’s own work demonstrates the direct application of AI-powered image analysis. He is developing tools that translate ai in agriculture images into actionable insights for farmers:

• Herbicide-Resistant Weed Mapping: Using cameras and AI software trained on the image repository, this tool maps the location of problematic weeds like Palmer amaranth. Knowing the precise location helps farmers target pesticide applications or implement crop rotation strategies more effectively, especially crucial for large farms spanning thousands of acres across multiple counties.
• Cover Crop Management: Analyzing images of mixed cover crops allows the AI to identify different species (e.g., legumes vs. grasses) and map their growth variation across a field. This information is vital for optimizing nitrogen fertilizer application for the subsequent cash crop (like corn), as legumes contribute nitrogen while others do not. Accurate mapping saves farmers money on fertilizer and reduces environmental risks associated with nitrogen runoff, such as groundwater contamination and algal blooms.

AI: The Missing Piece for Modern Agriculture

The integration of robotics (BenchBot), massive datasets (the image repository), and powerful computing (Grace Hopper) positions N.C. PSI at the forefront of agricultural innovation. These advancements are enabling the development of sophisticated ai in agriculture images tools that promise to make precision farming a widespread reality.

As growers face the dual challenges of feeding a growing global population amidst climate change and labor shortages, AI offers powerful solutions. While concerns about job displacement exist, AI also holds potential for easing agricultural labor shortages, particularly in labor-intensive sectors common in North Carolina.

“We’ve been talking about precision agriculture for decades,” Reberg-Horton concludes. “It’s really been aspirational for the most part… The AI revolution has been a missing piece that’ll help us get there.” By turning images into intelligence, N.C. PSI is helping sow the seeds for a more productive, efficient, and sustainable agricultural future.