Інтелектуальна масштабована система оцінки ваги ВРХ для задач точного тваринництва

Abstract

Abstract. This study presents an intelligent system for non-contact cattle live weight estimation and Body Condition Score (BCS) using computer vision. The solution integrates optimized 2D and 3D deep learning models to analyze animal geometry in real-time. By leveraging a Kubernetes-based microservices architecture, the system ensures dynamic scalability and superior GPU efficiency compared to traditional monolithic deployments. Experimental results demonstrate high precision, with weight estimation errors (MAE) under 18 kg and BCS accuracy exceeding 87 %, validating the system’s effectiveness for precision livestock farming.