Pig Face Recognition Application Using YOLO Algorithm and Transformer Model

The research revolves around the key utilization of the You Only Look Once (YOLO) algorithm for pig face detection, renowned for its real-time object recognition with impressive speed and accuracy. The increasing demand for intensified livestock farming necessitates accurate identification and traceability of animals, including cows and pigs. In this study, we propose a non-invasive, deep-learning-based biometric system for pig facial recognition. Our methodology involves several stages: firstly, we develop a ROS data collection module to capture facial information from ten pigs; subsequently, employing the SSIM method, we execute a preprocessing phase to eliminate highly similar images; finally, we employ an enhanced CNN image classification model (ViT), incorporating both fine-tuning and pretraining techniques for pig face recognition. Our proposed approach achieves an impressive 98.66% accuracy rate. In conclusion, smart farm technology, particularly employing the YOLO algorithm for pig face detection, holds immense potential for the livestock industry, especially in pig farming. Through image detection, movement analysis, and classification techniques, farmers can efficiently produce high-quality pigs with reduced labor requirements and minimized stress levels. This technology not only enhances pig grading and selection but also promotes animal welfare and sustainable farming practices. As it continues to evolve, it is expected to revolutionize the livestock industry and significantly contribute to the advancement of modern agriculture.