Automated Recognition of Cattle's Breed Through Computer Vision: A Case Study on the Oulmes-Zaer Breed

Abstract

Recognizing different cattle breeds accurately is important for farmers to prevent accidental crossbreeding and to maintain the purity of the breeds. The goal of this research is to simplify cattle breed classification through computer vision, with an emphasis on models that are suitable for mobile devices. We collected and curated a dataset consisting of 3,048 images, including images of the Oulmes-Zaer cattle breed, and relied on transfer learning to fine-tune various deep learning models on our dataset. The results show that the best-performing model is ResNet50, achieving an accuracy of 99.34% on the test set with a precision of 100<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>, recall of 97.41%, and F1-score of 98.70%. The quantized ResNet50 model requires 23.12 MB of ROM and 329.61 MB of RAM and has a latency of 108.33 ms, making it suitable for mobile and edge deployment.