Abstract
Background:
Recently, the depletion of livestock-sourced food supplies has become a major global challenge. This issue has led to rising raw material costs and increasing dependence on livestock imports from the surrounding region, prompting concerns about food safety and public health. Precise livestock characterization is important for ensuring traceability, managing disease, and preventing deceptive practices, thereby supporting the sustainable development of modern livestock industries. Recent studies using face and body patterns as the key biometric trait for identification have produced encouraging results. Furthermore, surveillance feeding patterns are essential for evaluating the health status and growth parameters of livestock. However, workforce deficiencies and strict biosafety protocols have made real-time livestock observation progressively difficult.

Aim:
This research was conducted to develop and evaluate a goat recognition model using the YOLOv11 framework. The main idea behind the development of the proposed model is to (i) automatically recognize individual goats with a high level of accuracy and (ii) implement a vision-based solution for automatic remote livestock monitoring with minimum human interference.

Methods:
This study evaluates the feasibility of using YOLOv11 to identify 7 known goats in a closed set under farm settings. After 10 training sessions, the model achieved an average mAP@0.5 of 93.7 ± 0.04%, with low variance between runs.

Results:
The proposed YOLOv11-based identification approach achieved overall mean ± SD values of 94.23 ± 0.18% for precision, 94.39 ± 0.07% for recall, 94.29 ± 0.10% for F1-score, 93.69 ± 0.04% for mAP@0.5, and 87.14 ± 0.09% for mAP@0.5:0.95 across 10 independent training runs. The class-specific mAP@0.5 values ranged from 91.11 ± 0.12% to 97.76 ± 0.11%, with Goat 1 showing the highest class-specific performance. The proposed framework effectively detects and monitors goats in real time.

Conclusion:
These findings support the feasibility of YOLOv11 for closed-set identification of known individual goats under the conditions represented in this dataset. Further validation using larger, multifarm datasets and previously unseen animals is required before establishing broader generalizability.

Key words: Automated goat identification; Livestock identification; Machine learning; Smart surveillance; YOLOv11.