The CattleTracs app makes photographing cattle a black or white deal – when the lighting, angle and distance are correct, the camera fires away.
PHOTOS COURTESY OF DAN DONNERT K-STATE RESEARCH AND EXTENSION.
Anyone who has uploaded photos to social media is familiar with the question: “Do you want to tag _ in this photo?” Sites like Google Photos can query every photo of a certain person based on just one snapshot. Human facial recognition is sophisticated, with many users from Facebook to airport security, to the FBI.
Now this technology is headed chute-side. Not to capture your workers’ pretty mugs – but those of your cattle.
The newly-released cell phone app CattleTracs has taken existing facial recognition technology and created an application
to identify cattle. The app was developed by parent company Black Hereford Holdings, Inc. out of Kansas City and created in
conjunction with research from Kansas State University.
CattleTracs was developed with the purpose of serving as a tool in disease traceability, although the owners and developers feel it has additional potential in the future.
KC Olson is a professor of range beef cattle nutrition and management at KSU and the lead of the CattleTracs development
team, a collaboration of researchers from animal science, veterinary medicine, computer science and engineering departments at KSU.
Olson says he got excited about the idea for the app because of the great need for rapid animal disease traceability. “Eartags fall out, paperwork gets lost – but an animal never loses its face,” he says. “The potential really caught my imagination. We have disease traceability now but it’s slow.” Those who remember the “cow that stole Christmas” with BSE in 2003 can appreciate the accuracy of a quick and accurate trace-back system as a way to limit loss of market access, value, and to avoid market crisis.
Use of the app is free to cattle owners and the application is completely voluntary – there are no reporting requirements tied to it. Currently CattleTracs has an enormously simplistic functionality and purpose.
Producers download the app on an Android or iPhone. The screen gives one option – to take a picture of a live cow (smart technology will not accept anything else, like a picture of your dog or a screen shot of a cow on a computer screen). After the first picture, the app equires two more from different angles, including at least one that has both eyes of the animal.
The app reads when the appropriate distance and angle from the animal are reached and automatically snaps the photos. After the third picture is captured, that animal is automatically and permanently uploaded to a database on an unalterable blockchain. Currently only two data points are captured with the photos – GPS location and date. Owner information or cell phone numbers are not contained or transferred in any way, and no other data points, such as sex or breed or animal age can be inputted as of now. If cell phone service is not available when the photos are captured, they stored until service is available then automatically uploaded.
Joseph Hoaglund, co-owner of Black Herford Holdings, Inc. had the idea for the app after seeing facial recognition used in Europe in the sheep industry. “I initially thought it would have an
application for the horse industry here, and we had a partnership with the American Quarter Horse Association going,” he says. However, the very nature of horses – they like to follow humans with their head and it’s hard to get different viewpoints – as well as the limited number of head per owner, made development challenging.
Hoaglund and his business partner, Brett Spader, transferred their efforts to cattle, particularly feeder cattle. The majority of food-chain cattle in the U.S. are routinely run through a chute at some point, greater numbers are at the same location, and the food safety factor fit the technology.
The key to a quality facial recognition program – be it human or bovine, according to Olson, is having a large database of input photos to train the neural network that runs the technology. The accuracy of the app – though not perfect – is encouraging. In groups of 10,000 animals the app can identify animals correctly 90 percent of the time. In a smaller proof-of-concept study at KSU on 1,000 head of yearlings the app was 94 percent accurate. Hoaglund says that although the technology will never be 100 percent accurate, there are two different uses of a facial recognition algorithm: “One is validation – is this the cow we think it is? And two is identification – Who is this cow?” Validation works on a much lower threshold of accuracy, says Hoaglund, and is the current use of the app. “Identification has to have a much higher level of accuracy, and that time will come but the we are not there yet.”
With identification the developers see opportunities for producers to include data points that are currently used in value-added EID programs, such as age, breed type, feed, handling methods, health records and other validated information used in
program cattle.
Although inputs are free, ultimately widespread use will come from upstream pressure from end users, envisioned to be groups like packers or grocery-stores. “It’s free to use, unless you want to get information back, then there’s a charge,” says Olson. Hoaglund is currently working with cattle producers in Brazil who are being sanctioned in Europe due to the unethical practice of cattle grazing destructed rainforest in the Amazon. “This application can help owners who are following the rules to prove the geographic location of their cattle, and prove they have not been unethically grazing in illegal areas.”
He is also in conversations with corporations he envisions could benefit from unalterable traceability information, such as packers and grocers. “The drive for this information will work its way up through the supply chain to people who are benefitting from that information: the packer, grocer and ultimately the consumer. If the consumer knows where their food comes from, they’re going to be healthier and safer than if they don’t know.” He notes a grocer in Europe that sells meat with a scannable bar code that pulls up a video of the meat animal’s previous life and growing conditions.
Interestingly, the facial recognition technology has been proven effective after the knock block in the packing house, adding a new and possibly more
effective option of tracing a carcass post-mortem that eliminates the need for tags and records.
Even for cattle not food-chain bound, is there potential this app could replace “the calving book”? Possibly, says Hoaglund.
“We’re living in the age of big data where everything we do is going to be entered and controlled through our cell phones,” says Hoaglund, “We are just starting to see the applications of artificial intelligence and nanotechnology in a way that is going to change the world as much as electricity did. In 1920 it wasn’t that we didn’t have electricity – we just didn’t have uses for it yet, like household lighting and vacuum cleaners and washing machines. Now, 100 years later, we have big data, and we are just starting to see the applications of it come about. It’s going to change our lives.”