July 1, 2022

Researchers using AI to detect chick distress calls in large poultry farms

Researchers from the City University of Hong Kong, together with their colleagues from the UK, have modified an AI application to detect chick distress calls in noisy large poultry farms, Phys.org reported.

The team's paper, which was published in the Journal of the Royal Society Interface, details the changes they made to an AI programme that was already on the market and how well it performed when put to the test on a real-world poultry farm.

Farmers look for ways to increase productivity as the world's population rises. In their latest endeavour, the researchers explored the possibility of using artificial intelligence to recognise and find distressed chicks housed in enormous farms.

Previous studies have demonstrated that animals typically call out to communicate their needs. For instance, a chick whose foot gets caught in a cage bottom will start chirping in a way that its mother will recognise as a cry for help. But in a big building full of other chicks, such chirps are hard to hear and isolate.

The new effort involved modifying the commercially available novel convolutional-neural-network-based model light-VGG11 to take out some of the extraneous parts and add some that are specifically designed to recognise baby chick distress calls.

They used 3,363 recorded distress calls and 1,973 regular calls to train the system to identify these calls. Then, they installed microphones in a barn that was home to thousands of live chicks in order to test the system. When it came to identifying the calls of distressed chicks, they discovered that the system was almost 97% accurate.

Before releasing their system for sale, the researchers want to make some adjustments. They point out that their system could be used to isolate problem areas and lessen instances of chick distress in addition to helping chicks who need it.

-      Phys.org