Characterization of behavioural changes associated with heat stress in dairy cattle using Computer Vision

Heat stress is a significant challenge faced by the global dairy industry as it impacts animal welfare, health and productivity. These impacts are becoming more widespread as dairy farming expands in Asia, but there is limited research examining heat stress in tropical and subtropical climates. Although numerous methods are used to detect heat stress in cattle, ranging from environmental measures to animal-based assessments, no definitive method has emerged for monitoring dairy cattle thermal status. Computer vision (CV) methods such as multi-object tracking (MOT) algorithms have great potential for use in livestock monitoring as they enable comprehensive identity-linked behaviour analysis with simultaneous tracking of multiple animals in complex environments. However, at present they are underutilized in livestock farming, with most models tested on benchmark datasets of pedestrians or vehicles.