The Insidious Side Effect of Using Facial Recognition Technology on Pigs

It could be cultivating more than just more efficient pork products.

Pigs in a field, with a face detection frame over one pig's face.
Photo illustration by Slate. Photo by Getty Images Plus.

Facial recognition technology, traditionally both marketed and feared as a way to exponentially enhance surveillance capabilities, may one day become integral for producing pork.

According to a report from the New York Times, major Chinese tech companies like Alibaba and are developing artificial intelligence tools to detect disease and keep track of individual pigs using facial recognition. China hopes the technology will make large farms more manageable, allowing it to consolidate and close smaller facilities. The government claims that the move would cut down on pollution.

Pig facial recognition would be invaluable to precision livestock farming, an animal husbandry practice most common in European countries and China that uses tracking technologies to maximize efficiency. Precision farming typically involves portioning out specific amounts of food to each animal depending on its age and weight in order to conserve resources. Animals are also constantly measured for growth and monitored for disease. Farmers can apply such techniques to pretty much any animal they may be raising, including pigs, chickens, cows, and sheep.

Farmers usually puncture pigs’ ears with radio-frequency identification tags in order to track them, though this practice has been criticized as cruel and time-consuming. Using facial recognition to monitor animals instead could be more humane, and might even help to detect disease and distress. Whether it’s ready for day-to-day use on a farm is debatable, however. Mark Hansen, a senior research fellow at the Bristol Robotics Laboratory, is one of the few academics who have conducted studies exploring ways to adapt the technology specifically to pigs. In a paper published in June, Hansen reports that he and a team of researchers were able to train a model to identify 10 individual pigs in a series of 1,553 images with a 96.7 percent accuracy rate. Not too shabby.

The researchers on Hansen’s team first collected images of the pigs’ faces by setting up a motion-sensing camera behind a water vessel. Whenever a pig approached to take a drink, the camera would snap a picture of its face. Deep learning algorithms then scanned the images and honed in on distinguishing features, namely the snout, wrinkles, eyes, and markings that often appear on the top of the head.

Facial recognition systems for pigs are not all that different from those used for humans. In fact, Hansen’s study found that a system trained only on human faces was able to distinguish between the 10 pigs with a 91 percent accuracy rate. “This indicates that many of the same features that the network has learned to be useful for discriminating human faces are also useful for discriminating pig faces,” the study reads, “[and] hints toward how a trained network for faces in one species may be transferable to other species.”

It should be possible, then, to modify facial recognition models built for humans to work on other mammals. Doing so could help to address some of the shortcomings of facial recognition algorithms for pigs, which often struggle with variations in expression and brightness. Hansen’s study notes that farms are “unconstrained” settings in which sunlight and dirt are not as controlled as they would be in a lab setting. Recognition algorithms trained on humans are more adept at adapting to different environmental factors that may affect accuracy. That same adaptability should carry over when reconfiguring a human face algorithm for pigs.

However, pigs have a number of characteristics that may pose challenges for facial recognition algorithms. Pigs typically age faster than humans, which would make it more difficult for a facial recognition system to track them through their growth. Facial recognition models tend to see a sharp decline in accuracy for pictures of a person taken more than seven years apart. Pigs reach maturity within a matter of six months, at which point they’re slaughtered. Pigs also attack one another from time to time, which can lead to injuries to the face, like a missing chunk of ear. And, as far as public research goes, it is currently unclear whether the recognition algorithms can adapt to different species of pigs.

But there’s a bigger problem. “I think the technology is there in terms of the algorithms and the accuracy in real world settings,” says Hansen. “But we have to prove … whether or not it’s actually useful to the farmers.” At the moment, the data that the algorithms spit out doesn’t mean much without other contextual information. “If [the technology] doesn’t translate into a management system that’s easy to use and robust, the farmers are not going to be willing to trial it,” says Hansen. “[Data] needs to be translated into ‘Pig No. 671 needs be taken out of pen No. 4 because biting behavior is about to start.’ ”

Convenience and practicality are not the only factors that facial recognition companies will need to consider when trying to convince farmers to use the tools. Ian Werkheiser, an assistant professor in bioethics at the University of Texas Rio Grande Valley, notes that some farmers in the EU have been resistant to the encroachment of precision livestock farming. Researchers in the EU have traditionally attributed this unwillingness to poor marketing, though Werkheiser posits a different theory. “It’s possible that farmers are actually correctly seeing that this would lead to an alienation of them from their jobs,” he says. He adds that relying on industrial automation, rather than humans, to handle pigs tends to cause more stress throughout the animals’ lives, which can lead to lower-quality meat products.

Werkheiser also worries about the farm tools subtly encouraging more complacency around human surveillance. “Companies that are making these technologies try to look for innocuous, even seemingly beneficial applications for the technology so that they can get people comfortable with it,” he says, adding that the proliferation of hobby drones for consumers may similarly be making the public more comfortable with their use by law enforcement and the military. “It’s very much in these companies’ best interest to be able to say, ‘[Facial recognition] is good for animal welfare. It’s good for health concerns.’ And so create more social comfortability with the technology.”

As we know, Chinese law enforcement is becoming increasingly reliant on facial recognition technology to keep tabs on citizens, particularly those who live in Xinjiang, home to the Uighur people. Facial recognition is also being aggressively introduced in the country’s stores, apartment buildings, and even public restrooms as a convenience perk. Marketing the technology as a quotidian farming tool may be yet another way to help people accept facial recognition as a quotidian component of modern society.

Future Tense is a partnership of Slate, New America, and Arizona State University that examines emerging technologies, public policy, and society.