Monitoring the body condition of dairy cows is essential for ensuring their health and productivity, but traditional BCS methods—relying on visual or tactile assessments by skilled personnel—are subjective, labor-intensive, and impractical for large-scale farms. To overcome these limitations, we present BCS-YOLO, a lightweight and automated BCS framework built on YOLOv8, which enables consistent, accurate scoring under complex conditions with minimal computational resources. BCS-YOLO integrates the Star-EMA module and the Star Shared Lightweight Detection Head (SSLDH) to enhance the detection accuracy and reduce model complexity. The Star-EMA module employs multi-scale attention mechanisms that balance spatial and semantic features, optimizing feature representation for cow hindquarters in cluttered farm environments. SSLDH further simplifies the detection head, making BCS-YOLO viable for deployment in resource-limited scenarios. Additionally, channel-based knowledge distillation generates soft probability maps focusing on key body regions, facilitating effective knowledge transfer and enhancing performance. The results on a public cow image dataset show that BCS-YOLO reduces the model size by 33% and improves the mean average precision (mAP) by 9.4%. These advances make BCS-YOLO a robust, non-invasive tool for consistent and accurate BCS in large-scale farming, supporting sustainable livestock management, reducing labor costs, enhancing animal welfare, and boosting productivity.

Accurate and scalable body condition scoring (BCS) is important for health monitoring and productivity management in precision livestock farming. However, manual scoring is subjective, labor-intensive, and difficult to standardize, while many automated methods are too computationally demanding for edge deployment in real farm environments. This study proposes EdgeBCS-YOLO, a lightweight object detection framework for real-time beef cattle BCS in unstructured farming scenarios. Built on YOLO11n, it combines Position-Sensitive Feature Fusion (PSFF), a Texture-Aware Star Module (TASM), an Efficient Grouped Detection Head (EGDH), and a Focal and Global Knowledge Distillation (FGD)-based distillation strategy. On a dynamic blurring dataset, EdgeBCS-YOLO achieved 90.8% precision, 82.7% recall, and 88.9% mAP@50. On the NVIDIA Jetson Orin NX Super, it achieved a model size of 3.95 MB, a system FPS of 33.35, and an average inference latency of 13.26 ms. These results suggest that it is a practical and potentially efficient solution for automated BCS on edge devices.