Abstract

Group housing of pigs is an increasingly common practice in modern agriculture, but it presents challenges in monitoring and ensuring the well-being of the animals and pig welfare. In this study, we propose the use of Convolutional Neural Networks (CNNs) for the behavior classification of group-housed pigs. We collected a dataset of pig behavior images, annotated with various behaviors including drinking, eating, excreting, sleeping, and standing-up behaviors. A CNN model was trained to classify these behaviors, with a focus on identifying stress-related behaviors. The results indicate that YOLOv8 exhibits competence in the classification of group-housed pig behavior, achieving an accuracy rate of 81.7%. Furthermore, it showcases noteworthy proficiency in classifying the "drinking" category with a precision of 96.7%. The model showed promising results to quickly detect and address pig interactions, offering a potential tool for real-time monitoring of pig behavior, thereby promoting better animal welfare.