Among the three behaviors analyzed, Tail Raising emerged as the most reliable indicator of imminent calving. It showed a clear increase about 5 to 6 hours before calving, aligning with prior studies and supporting its role as a practical early warning sign. Head to Abdomen behavior, while more variable, also contributed meaningfully to prediction in certain individuals. Although this behavior did not display a strong population-wide trend in statistical modeling, its temporal changes were effectively captured by machine learning algorithms, suggesting its potential in personalized monitoring. Lying-Standing Transitions increased gradually throughout the prepartum period but lacked a sharp temporal threshold, making them less useful for short-term calving alerts.
Looking ahead, future studies should consider expanding video coverage to include feeding and watering areas, which may capture additional behavioral cues relevant to calving. Incorporating new indicators—such as pacing, total movement, or circling—could also enrich behavioral profiles and enhance model sensitivity. Moreover, collecting background information like parity, health status, or genetic traits may help account for individual variation and improve prediction accuracy. Technical improvements, such as analyzing differences between daytime and infrared footage, can also guide better camera setup and model calibration. Finally, increasing the sample size will be essential for building more generalizable models that can be applied across diverse farm environments.
This study demonstrates the feasibility and value of using markerless computer vision tools to monitor fine-scale behaviors in dairy cows. By combining DeepLabCut with machine learning, it provides a low-interference, scalable solution for calving prediction. Tail Raising stood out as the most informative behavior, but the integration of multiple subtle cues enhances the overall robustness of the model. The approach offers a promising alternative to sensor-based systems and contributes to more efficient, welfare-focused calving management in modern dairy farming.