Artificial intelligence can spot baby chickens in distress | Science

Chickens make more noises than most of us realize. They cluck when pleased, screech when frightened, and sing “buk, buk, ba-gawk” when laying an egg. Their chicks also vocalize, and they can vary that simple sound to signal pleasure or distress. Now scientists have developed an artificial intelligence (AI) program that automatically identifies these SOS calls, an advance that could help farmers save thousands of young lives — and millions of dollars in agricultural labor.

“The results are an important next step towards an indicator for flock welfare,” said Bas Rodenburg, an animal welfare scientist at Utrecht University who was not involved in the study. The work could even change public attitudes toward factory farms, he says. In general, the the general public prefers to eat chickens— and farm animals in general — from empathetic producers who care about the welfare of their animals, other researchers have shown.

Early in their lives, chicks utter cries for help –high, repetitive beeps-to attract the attention of their mother hen, on whom they rely for warmth and food. She responds with food calls and shows the chicks where to forage. But in a commercial hen house, chicks call when they feel uncomfortable, socially isolated, or hungry. Answering these calls can be the difference between life and death: Ignored chickens are more likely to lose weight and die prematurely† Animal welfare scientists have tried to develop automated methods to help farmers better recognize these situations.

To enhance these efforts, researchers from the City University of Hong Kong recorded the sounds of chickens housed at Lingfeng Poultry Ltd., a major poultry producer in China’s Guangxi province. The birds are kept in stacked cages (three cages per stack and 13 to 20 individuals per cage), with approximately 2000 to 2500 hens in each house.

A. Mao et alJR Soc. Couple

Over the course of a year, the researchers surveyed the environment, picking up everything from natural farm sounds like workers scrubbing barn floors to the distress calls of chicks. They then transformed all these sounds into sound images known as spectrograms and used the images to train a kind of AI program called deep learning. Similar programs have been developed to emotional states of cows on dairy farms

Using the recorded sounds of the barns and sounds created in real time in a live demonstration, the algorithm can quickly and successfully identified 97% of emergency calls while the chickens were making them, distinguishing it from other chicken noises and from the general barn noise, the team reports today in the Journal of the Royal Society Interface

Still, “more work is needed” to make the research “commercially viable in the real world,” said Marisa Erasmus, an animal behavior and welfare scientist at Purdue University who was not involved in this study — a fact that the Hong Kong scientists acknowledge, too. Nevertheless, because the approach worked in real-time when chickens made emergency calls, it’s a big step forward, she says. It “brings researchers one step closer to automatically tracking the health and welfare of farm animals.”

For example, she and Rodenburg can envision an alert system in a large commercial farm that alerts workers to a specific cage where a chick is in distress so they can provide needed and timely care. And that could lead to a future with many more happy clocks.

Leave a Comment

Your email address will not be published. Required fields are marked *