Contextualized Small Target Detection Network for Small Target Goat Face Detection

SIMPLE SUMMARY: Goat identification is highly demanded in modern livestock management, and sheep face detection is an important basis for goat identification, for which we developed a new computer model that overcomes the challenges of unclear images, small targets, and low resolution. By considering the surrounding details and combining different features, our model performs better than existing methods in detecting goat faces. We used various evaluation metrics to measure its effectiveness and found a significant improvement in accuracy. The results confirmed that our method successfully addresses the difficulty of detecting lamb faces. This study has important implications for the development of intelligent management systems for modern livestock farms to better identify and monitor goat for improved animal welfare. ABSTRACT: With the advancement of deep learning technology, the importance of utilizing deep learning for livestock management is becoming increasingly evident. goat face detection provides a foundation for goat recognition and management. In this study, we proposed a novel neural network specifically designed for goat face object detection, addressing challenges such as low image resolution, small goat face targets, and indistinct features. By incorporating contextual information and feature-fusion complementation, our approach was compared with existing object detection networks using evaluation metrics such as F1-Score (F1), precision (P), recall (R), and average precision (AP). Our results show that there are 8.07%, 0.06, and 6.8% improvements in AP, P, and R, respectively. The findings confirm that the proposed object detection network effectively mitigates the impact of small targets in goat face detection, providing a solid basis for the development of intelligent management systems for modern livestock farms.