AI 'Weighs' Cattle With Just a Glance
Source PublicationPLOS One
Primary AuthorsHossain, Ferdaus, Islam et al.

Accurately gauging a cow’s weight is vital for modern farming, but traditional methods are labour-intensive and often imprecise. While computer vision offers a solution, older machine learning models struggle to interpret the complex visual information in an image, leading to less accurate predictions.
Now, researchers have introduced CattleNet-XAI, a new framework built around a custom Convolutional Neural Network (CNN). A CNN is a type of deep learning AI designed to process and understand visual data, much like our own brains recognise objects. To improve its performance, the system first enhances the cattle images using techniques like normalisation.
When tested against other models, including established methods like Random Forest, the new custom CNN proved superior. It achieved an impressive accuracy, with an average error of just 18.02 kg. Crucially, the framework also includes features for explainability, allowing researchers to gain insights into the model’s decision-making process. This technology promises a more efficient and precise approach to livestock management.