Answers are grounded in this paper and cite their sources.
Survey evaluates the application of large-scale foundation models in improving precision livestock farming, suggesting significant advancements and challenges.
Precision Livestock Farming integrates advanced technologies such as the internet of things (IoT), artificial intelligence (AI), big data and machinery automation into traditional animal husbandry to enable data-driven, automated, and precision farming. Operating within a perception-decision-action hierarchical framework, this paradigm facilitates the precise regulation and dynamic optimization of production processes, thereby enhancing animal welfare, productivity, and economic efficiency. Despite these benefits, the development of precision livestock farming faces challenges, including heterogeneous data types, the scarcity of large-scale annotated datasets, and the complexity of multimodal data fusion. The advent of large-scale foundation models (LSFMs), characterised by their massive parameter count and pre-training on vast datasets, offers a transformative paradigm for addressing these persistent challenges. In this review, we present a comprehensive survey on the application of pre-trained LSFMs in precision livestock farming, offering a structured synthesis of recent advancements. This survey first reviews the recent advances in LSFMs, with a particular focus on multimodal architectures. Subsequently, a novel taxonomy is introduced, categorizing LSFMs applications across six core functional subsystems of precision livestock farming: inventory monitoring, growth monitoring, status monitoring, intelligent decision-making, animal feeding management and animal health management. For each subsystem, existing methodologies are systematically analyzed. The survey culminates in a critical discussion of current limitations and outlines promising directions for future research. • A novel application-oriented taxonomy is proposed to structure LSFMs across six core operational subsystems of smart livestock farming. • A comprehensive cross-modal analysis is conducted to evaluate LSFM capabilities in processing and fusing text, image, video, and audio data. • Embodied AI models are integrated into the husbandry context for the first time, enabling action-oriented perception and autonomous robotic support. • An extensive collection of livestock-specific multimodal datasets is curated to facilitate standardized benchmarking and future model development.
No takes yet. Share an insight, caveat, or question.
Tian et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: