Artificial Intelligence on Farms: Sheep Breed Classification Using Computer Vision

Abstract

As the challenges of population explosion, food security and climate change become more pressing, there is a dire need to produce more food. Since animal husbandry is a crucial part of agriculture, production per farm has to also increase substantially over the next few years. This is especially important for developing countries where the introduction of technology and automation can remove drudgery, improve efficiency and reduce labour. An important part of this is the use of images for automatic animal monitoring. This research was therefore undertaken to classify four breeds using three artificial intelligence algorithms; K-nearest neighbours, Support Vector Machines and Convolutional Neural Networks (CNNs) which were evaluated and ranked. We thus developed three deployable models for image classification among which pre-trained CNNs had the highest accuracy of 0.90. We conclude that image classification therefore is useful for automatic animal classification and monitoring. This would help improve the production potential of farms through the automation of farm tasks.