Call me naïve, but I always assumed when I saw a combine out working some corn the giant, shiny steel behemoths were sucking up 100% of the grain (or something close to it) as it devours whole swaths of dried down crops during fall harvest.
I learned this week that is not the case. The University of Georgia says that grain loss of up to 10% or more is typical, and that farmers should happily accept grain loss in the 2 to 4 percent range as normal operating procedure.
That really brings into question the whole “feeding 9 billion by the year 2050” thing. As one Iowa farmer told me as we perused the components of farmwave’s Harvest Vision AI tool laid out on a folding table at The Smoker Farm in Wanatah, Indiana, “think about how many we could feed if we were just a little bit more efficient.”
He’s certainly got a point there.
What is farmwave?
The farmwave system applies advanced artificial intelligence (AI) and Machine Learning (ML) algorithms to quantify in real time just how much grain is being lost from the header and out the back of the combine.
The companies’ founder and CEO, Craig Ganssle, is completely transparent about his expertise: we’re not hardware people, he says, we’re data and AI people. Most of the components that make up his proprietary system are what could be considered “prosumer,” or off the shelf electronics.
The true power behind farmwave, according to Ganssle, is in its AI image recognition capabilities.
Before the days of farmwave, farmers would stop the combine, get out and climb down the ladder, and look around behind the combine to see how much grain was scattered about. Or drop pans can be used to capture and quantify lost grain just behind the header.
According to Google, Pioneer Seeds says that 20 wayward corn kernels within a 10 sq. ft. area behind the combine means the farmer is losing roughly one bushel of grain per acre.
If you’re an early adopter of the farmwave system you don’t have to guess how much grain you’re losing. The software can tell in real-time by taking a picture every three seconds and running AI over the images. Just like how a smart sprayer has downward facing cameras to identify weed from plant, the farmwave cameras distinguish wasted grain from trash, dust, or whatever else might be getting chopped up and spit out the combine.
As it stands now, farmwave users can set targets for acceptable loss and unacceptable loss in the product’s software, and a graph on the farmwave farmbook – a 12” ruggedized Dell tablet capable of running AI algorithms at the edge (no continuous connectivity is required of the system at any time) will alert the operator to stop and make combine settings adjustments. That could include changing the header height, chaff speed, or even the combine’s overall ground speed.
The companies’ founder and CEO, Craig Ganssle, says by 2026 the product will have some level of automation built into it (through the ISOBUS terminal) that will make those adjustments automatically based on what the cameras are seeing.
The present iteration of farmwave represents quite the leap forward from the product’s prototype stage. Ganssle recalls six years ago getting the initial system got off the ground with “suction cupped GoPros” that required someone to pull the SD cards post field pass and upload the images to see what the cameras were seeing.
Today farmwave has ruggedized cameras with stereo pair LiDAR sensors that can be installed basically anywhere on the combine that the farmer wants to monitor for grain loss. According to Ganssle, the LiDAR sensors are integrated to associate depth with each image the camera takes, and that “(we have) yet to tap into this resource (LiDAR), making these cameras future proof.” All images are GPS referenced as well, since the farmbook has its own GPS chip inside.
The images are uploaded into farmwave’s Google Cloud environment for storage once the combine hits WiFi. That feature can be turned off if the farmer doesn’t want to share the data and just wants to visualize it in real time from the tablet, and many farmers do just that, according to Ganssle.
Another cool feature to note is a user can build Harvest Loss Maps and upload those into agronomic analysis software to better understand where harvesting issues are happening in the field, without having to connect to the cloud. It can all be done on the portable farmbook and exported as a standardized ag data format.
If a farmer does wish to share data with the cloud, the images are stored there online (the farmbook has limited storage capacity, like any machine) and remain the property of the farmer. They may be used to train farmwaves’ AI and machine learning algorithms.
Farmer tests out farmwave
Jacob Smoker, whose father Greg owns The Smoker Farm just off Interstate 30 in Wanatah, Indiana, is one of several farmers pilot testing farmwave across his own ground.
By adjusting his fan speed and the header itself during testing, the reduction in grain loss quickly covered his upfront cost for the system.
Depending on how many cameras the user wants, the farmwave system costs between $12,000 to $16,000 upfront and then a yearly subscription fee of $2,500. The system can be installed on a combine in about two hours if you’re good at wrenching on equipment and running wires, even on older combines from the 1960s.
While its price tag might shock some, that same Iowa farmer told me he wouldn’t blink an eye at plunking down that much to get his hands on farmwave’s technology. In fact, he’s also thinking about becoming a regional dealer and selling some systems to his neighbors.
One of the differentiating points that stood out is farmwave’s inclusion of farmers themselves in the product’s development cycle. Although Ganssle did not grow up on a farm, the U.S. Marine Corps military veteran has deployed a solid and experienced advisory board of tech-focused farmers (well known podcast host and ag tech advocate Nathan Faleide among them) to “tell us what the system needs to do (for them) and we’ll get working on it.”
Want to learn more about farmwave? Check out its FAQ page for videos from Ganssle breaking down the system.
