Computer Vision in Livestock and Poultry Farming: A Quantitative and Translational Review of Research Trends and Real-World Impact

• Quantitative synthesis of 191 CV studies in livestock and poultry farming • Behaviour monitoring and identification dominate current research across species • Architectural innovation and benchmarking drive methodological advancement • Limited welfare validation and productivity evidence beyond visual inference • Data openness and reproducibility remain key challenges for real-world adoption Computer vision has emerged as a key enabling technology in precision livestock and poultry farming, with several studies exploring its various applications. However, the extent to which their technical advances translate into biologically measurable welfare assessment, management decision support, and real-world impact remains insufficiently synthesised. This review provides a structured analysis of 191 peer-reviewed articles and conference proceedings published between January 2020 and August 2025, covering computer vision applications across seven major livestock and poultry species. It synthesises publication trends, application domains, methodological advancements, translational validation practices, and practical deployment. Studies were retrieved from Web of Science, Scopus, IEEE Xplore, and ScienceDirect, with screening and coding conducted using predefined criteria. Quantitative analysis reveals a strong research emphasis on behaviour monitoring and individual identification, particularly in cattle, pigs, and chickens, while small ruminants and less-studied poultry species remain underrepresented. Methodologically, the field demonstrates increasing architectural sophistication and widespread benchmarking. However, public dataset availability (12.04%) and open-source code sharing (5.76%) remain limited, constraining reproducibility and cross-study comparability. Evaluation of translational validity shows that only 15.71% of studies incorporated welfare or health validation beyond visual inference (validation against independent reference measures, including veterinary scoring, physiological indicators, or laboratory diagnostics). Fewer than 5% reported quantitative evidence of productivity or management impact, even among studies reporting on-farm deployment. These findings highlight a persistent gap between technical performance metrics and biologically or managerially measurable outcomes and identify key opportunities for advancing computer vision toward biologically grounded and managerially relevant solutions in precision livestock and poultry farming.