A cow’s vocalisation changes when it enters its oestrous phase. The changed vocalisation is useful to attract bulls in the vicinity and to ensure reproductive success. Though the change in vocalisation is subtle, experienced farmers have learned to recognise it and take steps to either bring in a bull or to bring in a vet to artificially inseminate the cow in oestrous. But, in the large scale farming of cows in the diary industry, people who are not so experienced are employed. That is when artificial intelligence can be helpful.
H M Rohini and S Prabhavathi at RYMEC, an engineering college in Ballari teamed up with Doddabasappa Angadi at the Sri Sathya Sai University for Human Excellence, Kalaburgi, Karnataka to teach three different convolutional neural network architectures to differentiate between the call of cows in oestrous and cows that are not in oestrous.
LeNet, a lightweight convolutional neural network, originally designed for picture categorization, captures key acoustic features. VGG 16, a complex model with many convolutional layers, is good at optimising feature-extraction. The ResNet model is good at improving accuracy. The researchers combined the strengths of the three models by creating an ensemble. The ensemble model, when tested, showed more than 96 percent accuracy in distinguishing the vocalisations of cows in their reproductive phase.
The dairy industry in India needs to step forward and evaluate the usefulness of the ensemble model in real world settings.
Letters in Animal Biology 06(2): 139 – 145 (2026);
DOI: 10.62310/liab.v6i2.391
Leave a Reply