In the “Year of the Pig” – Here’s Facial Recognition for Pigs

China’s endeavours in using facial recognition for the surveillance of its populace is well documented, as is the fact that the country boasts the densest network of AI-driven facial recognition cameras in the world. Less well known, perhaps, is that the country’s top tech companies are now turning their facial recognition cameras at a completely different target: pigs!

In the Year of the Pig, well-known e-commerce giants such as Alibaba, and Tencent are developing AI-based Smart Agriculture platforms which they believe will lead to improved agricultural efficiency, particularly with respect to pig rearing.

Pork is by far the most popular meat dish in China and more than half of the world’s pigs are reared and consumed in the country. However, pig farming in China is woefully inefficient compared to western countries and Chinese tech companies see opportunities to apply AIoT solutions to remedy this. For instance, Alibaba is using its so-called “Agriculture Brain,” an AI platform that uses AI-supported facial and speech recognition plus other AIoT technologies, to help farmers monitor pigs in real time.

Pig facial recognition works in a similar way to human facial recognition, recording details of the pig’s eyes, ears, snout and bristles. Like humans, pigs do not all look alike! Audio recognition detect symptoms such as coughing and breathing difficulties while infra-red technology is used to detect the pig’s temperature. In addition to identifying the pig, the facial recognition camera also monitors the pig’s movements and can detect whether the pig is becoming lethargic or not eating. Each pig has its own archive file, which includes key details such as its age, weight, breed, exercise frequency and its feeding habits, often a good indicator of a pig’s health.

Using all this data, the AI platform continuously monitors the pig and triggers an alarm if it suspects the pig is unwell so that early diagnosis can be made. In 2018, the outbreak of African swine fever virus resulted in nearly a million pigs being slaughtered to contain the disease, so there is hope that AI-based systems can help prevent similar contagion in future.

Counterpoint Research is aware of several big pig farms in China that are currently using these AI-based systems. But how accurate can pig facial recognition really be and how do the companies overcome some of the obvious challenges? For instance, how do they get a pig to stand still in front of an AI camera long enough for it to be identified? Pigs live in a dirty environment and can easily soil their faces, which would make it more difficult for a camera to identify them.

Start-up Yingzi Technology, one of the first Chinese companies to roll out a pig facial recognition system, is trialling its system on a farm with 3,000 pigs. The system works by continuously scanning each pig’s face using a smartphone, which records the pig’s distinguishing features (see the picture below). The data is then analysed by an app on the phone using deep learning algorithms and uploaded to an online database. The company claims that system matches and updates the individual pig’s profile in just a few seconds and documents details such as its ID number, its breed, birth date, weight, gender and genetic composition.

Furthermore, Yingzi claims that identification accuracy is more than 98 per cent and that a pig can be identified even if it is moving in a herd. However, there is no scientific evidence to back these claims as government rules to fight the swine fever prevent outsiders from visiting pig farms to see the technology in action, so claims by the companies cannot be independently verified at present.

As well as improving pig rearing efficiency and animal healthcare, another objective here is to improve food traceability. China has a history of food safety scandals over the years, from melamine-tainted eggs, to smuggled and out-of-date frozen meat, to crops tainted with heavy metals. Ultimately, Yingzi wants to develop a system whereby customers are able to scan a PR code on the food packaging which immediately informs them of the full provenance of the pig.

Counterpoint Yingzi Technology’s facial recognition for pigs

Exhibit: 1 Yingzi Technology’s facial recognition for pigs

Despite serious doubts about some of the accuracy and reliability claims made by these companies, the use of AI-based systems in livestock farming is expanding worldwide and are being used to improve animal husbandry more universally. For example, US global food company Cargill recently partnered with Irish machine vision company Cainthus to develop facial recognition for cows. Clearly, further studies are required to demonstrate whether these systems are able to consistently identify an animal under all circumstances, and it is likely that it will take a few years for the technology to be perfected. However, one thing is sure: facial recognition is expanding rapidly from the human world to the animal world.