Farmwave helps farmers stop crop loss with Coral

In the quest to feed a growing global population, few actions are more effective than cutting food waste. According to Project Drawdown, around a third of the food farmers produce never make it to the table. For example, growers lose as much as 10 percent of their grains or beans at harvest.

That’s badly needed money (and food) left behind. It also contributes about eight percent of global greenhouse gas emissions, compounding another major problem.

“We actually grow more than enough food to feed everybody on the planet,” says Farmwave founder and CEO Craig Ganssle. “But waste is a very big problem, and it starts out in the field.”

The lost crop hits small farmers particularly hard. “We are all just trying to survive,” says Jake Smoker of the Smoker Farm in Wanatah, Indiana.

Now, however, farmers have access to an AI-powered tool to help them tally potential losses during harvest, allowing them to make adjustments in real-time. Smoker calls it a game-changer. “A small change in settings, and I can save literally thousands of dollars,” he says.

The tool, from Farmwave, includes custom-developed ML models, a harvester-mounted box with cameras, an in-cab display, and on- device AI acceleration from Coral.

Tracking loss

Farmers magnetically attach Farmwave devices to their combine harvesters, no tools or time-consuming retrofits needed. Once the machine is underway, the app gives the farmer in the cab real-time updates about how efficiently the machine gets crops from plant to grain bin. Armed with that information, the farmer can make adjustments on the fly, including to the angle and speed of the rotor and sieve openings.

Growers have long known that the right adjustments to harvesting machines in response to conditions in the field can make a big difference in yield. However, conventional, manual methods for estimating the efficiency of a given harvest require stopping the combine and, in the case of corn, painstakingly counting individual kernels.

That’s not feasible for many farmers, who measure the value of their half-million-dollar combines in minutes spent running them.

“All in a world where we’re hoping to make $50 an acre, we could pick up another $30 because we had the machine better adjusted,” says Dan Lucas, Precision Agronomist at AgriVision Equipment Group, a farm equipment dealer in Iowa.

Harvesting a revolution

Ganssle founded the company that became Farmwave in 2007 after stints in the Marine Corps and Verizon. A farm equipment developers conference in 2013 sparked the inspiration for Ganssle that image recognition could help solve significant problems for agriculture. “It started with crop scouting,” says Ganssle of his initial investigations, “looking for disease and pests in the field.” The company began to focus on crop loss in 2019. “We were talking to some people in the machinery world, and they said, ‘If you can solve for harvest loss, that would be a game-changer across the industry.’”

Farmwave launched its loss analysis system in time for the 2020 harvesting season. Even before harvest began, reviews were glowing. “Something that can monitor loss accurately, report to me, and do it in a way that is simple to use and accurate is incredible,” Smoker says. “Farmwave is going to change the landscape of agricultural technology with this type of technology; I have no doubt in my mind.”

Combines are slow, typically covering 12 acres in an hour. Even so, that’s fast enough to make cloud-based AI infeasible for tracking a harvest in real-time. That’s why Farmwave’s devices must have AI onboard to make a difference for farmers.

The Farmwave devices, able to withstand vibration, dust, and temperature extremes in the field, use GoPro cameras to watch the grain passing from plant to grain bin. AI-powered software counts individual kernels of corn or soybeans and calculates the farmer’s loss per acre given current operating conditions.

Three units—each with two Coral USB Accelerators and two Raspberry Pi processors on board—go on the outside of the combine. Two are on the front of the machine and one at the back. In each unit, a Raspberry Pi connected to two Coral USB Accelerators handles picture-taking and processes the AI images. The Pi is networked with a second one for data collection and communication. A tablet in the cab keeps the operator updated on the harvest’s progress, based on consolidated data from the three units. Given the challenges of connectivity on many farms, the system is designed to do all of its work locally, with no cloud processing.

“There are no wires,” Ganssle explains. “Everything is battery-powered, magnetic-mount, and wireless.” In addition, the model for a given crop comes pre-loaded, no training required by the farmer. It’s this “secret sauce” that Ganssle and his predecessor company have spent seven years developing. He says he appreciates the “collaboration of AI tools and Google Cloud Platform working together. It’s a seamless workflow for us. We found what we needed in the Coral devices.”

Farming’s AI future

Farmwave enables individual farmers to subscribe to loss tracking as a service. But it is also working with major equipment makers such as AGCO Corporation and John Deere to build the system into future combine models. Other plans include enabling the system to watch for pests, blight, crop damage, and more.

Ganssle and the Farmwave team aren’t the only ones betting on AI to help boost harvests. Julieta Ross, founder of startup Okee Labs, is developing AI to help the 87 percent of farms around the world that cover under five acres. “There is a difference between a large farm that grows corn in 10,000 acres versus a small-holder, urban farm,” she says. Those smaller farms, many little more than garden plots, typically support multiple types of crops and often aren’t harvested with machines, presenting extra challenges for an AI-assist. The Okee Labs system will stand up in a field, watching for signs of blight, dehydration, and even predict future conditions based on local weather patterns. The company’s prototype runs TensorFlow on a Raspberry Pi for local AI processing, allowing it to work in areas without access to the cloud.

All of which adds up to democratizing AI for agriculture. For farmers like Smoker, it can’t happen soon enough.

Read more about the ways in which on-device ML can enhance agriculture operations, as well as other industries. If you’re interested in how Coral can help you build your next product, please contact us.