Artificial intelligence

Alibaba uses AI to speed up detection of pregnant pigs seven times, boosting efficiency of China’s hog farms

  • China has 700 million pigs, representing about half of the world’s total
PUBLISHED : Wednesday, 19 December, 2018, 6:01am
UPDATED : Wednesday, 19 December, 2018, 2:36pm

Chinese pig breeders have accumulated years of experience to determine if a sow is pregnant, based on careful observation of the animal’s behavioural changes over a 21 day period – a task made more challenging because each farmer has to keep tabs on up to 80 pigs at a time.

Now there’s a way to do it in just three days, thanks to artificial intelligence.

Using facial-recognition cameras, Alibaba Group Holding's cloud unit said it has developed an algorithm that can accurately diagnose pregnancy in pigs by observing changes in their behaviour, physical appearance and eating patterns after mating.

The solution, which can also detect a failed pregnancy as early as the third day after mating compared to 21 days the conventional way, is designed to increase the number of newborns on pig farms, according to Alibaba Cloud. The solution is expected to be put to use in some of Alibaba Cloud’s partner farms in southwest China’s Sichuan province early next year.

China has 700 million pigs, representing about half of the world’s total. In recent months a number of internet companies have launched ventures to revolutionise the country’s pig farming industry. goes the whole hog with facial recognition for agriculture

In November, JD Digits, a unit of China’s second largest e-commerce giant, announced its foray into the agricultural industry with a package of digital tools including machine vision, AI-enabled analytics, robots and internet of things technologies, to help the country’s farmers raise better animals at lower cost.

The so called pregnancy test algorithm is the latest development of the cloud unit’s ET Agricultural Brain, a digital tool aimed at boosting agricultural efficiency, crop yields and income for Chinese farmers by enabling them to make better use of big data. In some of the farms in Sichuan the system is already being used to track the growth of individual pigs and monitor their health condition.

A commonly used method to determine if a pig is pregnant is to observe the female’s behaviour when a boar is present about 21 days after mating. Sows that sleep with all four legs in the air, and those that prefer to stand still rather than move around, are also more likely to be pregnant – signs that can be detected by the AI tool.

“Apart from the observation of the behavioural changes, a more thorough check is quite time and energy consuming, requiring help from two to three people,” said Deng Qiang, who works at a pig farm under Alibaba’s smart agricultural project in Sichuan.

He said one person usually monitors the growth of 80 pigs simultaneously, so there was a lot of room for new technologies to increase efficiency.

“Even if the solution can only tell which pig is more likely to be pregnant, that is a big help for us,” Deng said,

New York-listed Alibaba is the parent company of the South China Morning Post.