Real-time gender classification of Garut sheep within a small flock using deep learning

¹ Department of Electrical Engineering, Politeknik Negeri Bandung, Bandung, Indonesia
² Department of Information Technology and Communication, Politeknik Kuching Sarawak, Sarawak, Malaysia
³ Department of Mechanical Engineering, Politeknik Kuching Sarawak, Sarawak, Malaysia
⁴ Department of Electrical Engineering, Politeknik Kuching Sarawak, Sarawak, Malaysia

^* Corresponding author: rida_hudaya@polban.ac.id

Abstract

Garut sheep are a unique Indonesian breed with both agricultural and cultural importance. Accurate gender identification is crucial for breeding management, health monitoring, and behavioral analysis. This study proposes a hybrid deep learning approach to automatically classify the gender of Garut sheep using video data. The method integrates You Only Look Once version 8 (YOLOv8) for real-time sheep detection with a Convolutional Neural Network (CNN) for gender recognition. Publicly available videos were collected, annotated, and preprocessed to construct a custom dataset. YOLOv8 was applied to detect sheep in frames, and cropped regions were then passed to the CNN for classification. Experimental results show that the combined YOLOv8–CNN framework performs effectively in distinguishing male and female Garut sheep under varying lighting and motion conditions. The system achieved an accuracy of 89.6% while maintaining real-time performance at 25–30 frames per second (FPS) from CCTV video input. These results demonstrate the promise of computer vision for intelligent livestock monitoring in precision agriculture. Nevertheless, the study remains limited by the small dataset size, which may affect generalizability. Future work will focus on expanding the dataset, capturing more diverse conditions, and validating the system’s robustness on larger flocks.