Fig. 1: End-to-end framework for video-based cattle behaviour detection and digital-twin integration.
From: Video-based cattle behaviour detection for digital twin development in precision dairy systems
Continuous video streams from barn cameras are processed through a pipeline comprising (1) object detection and multi-cow tracking (YOLOv11 + ByteTrack), (2) behaviour classification using deep spatiotemporal models (SlowFast / TimeSformer), (3) behavioural data translation into structured activity logs for nutritional modeling via NRC equations, and (4) dynamic synchronization with a Unity-based 3D digital-twin environment for visualization and decision support in precision dairy management.
