Description of the research group

• About the group

  The Computational Ethology and Precision Animal Welfare Research Group pairs computer vision and machine learning with applied ethology to deliver automated behavioural phenotyping from continuous sensor data. We are a cross-faculty group at NMBU, spanning Veterinary Medicine, Biosciences, and Chemistry, Biotechnology and Food Science. Where conventional welfare assessment depends on periodic observation, we decode internal states continuously, measuring positive affect alongside distress. We translate behavioural time series into validated decision support for veterinarians and producers, extending precision livestock farming from productivity metrics to welfare outcomes. Because the sensing and phenotyping pipeline is species-agnostic, the same methods transfer to laboratory and companion animals. The group is led by Professor Andrew Michael Janczak of the Section for Animal Welfare, Epidemiology and Population Medicine at the Faculty of Veterinary Medicine.

• Focus areas

  We study how animals perceive, cope with, and are affected by their physical and social environment, and how that experience can be read from behaviour, quantified continuously, and turned into better decisions for animals in human care. We work on core welfare challenges including stress, fear, pain, lameness and damaging behaviour, and on promoting positive emotional states. Our main focus is on assessing animals at the individual level.

  We work across the full arc from mechanism to computation and practice.

  At the mechanistic level, our expertise spans the neural control of behaviour and physiology (including the functional neuroanatomy of emotional and cognitive responses to environmental stimuli) to nociception and pain perception, the gut-brain axis, and pharmacological and physiological correlates of welfare states.

  At the behavioural level, we draw on deep roots in applied ethology. This includes behavioural development, environmental enrichment, play and the study of positive emotional states, alongside the detection of distress and damaging behaviour.

  At the computational level, we develop computer vision, machine learning and digital phenotyping methods that convert continuous behavioural data into interpretable, confidence-rated and cross-validated indicators.

  At the level of practice, we carry these methods through to validated, user-facing decision support for farmers, veterinarians and animal keepers. This is supported by an in-house capability for digital product development, deployment and operations, where privacy, ethics, human oversight and regulatory compliance are designed in from the start.

• Across species

  A defining feature of our group is our reach across species and across levels of biological organisation.

  Our work ranges from fish (where research on the neural and physiological basis of behaviour supports both improved welfare and the sustainable use of farmed and wild fish resources) via poultry, pigs, cattle and rabbits, where we address stress physiology, cognitive development, robustness, lameness and the prevention of damaging behaviours such as feather-pecking and tail-biting. In addition, we cover companion animals, laboratory animals and other species in human care.

  This breadth is not incidental. It reflects a deliberate commitment to behaviour-based welfare frameworks where the principles generalise across managed populations. This gives empirical substance to a genuinely species-agnostic methodology, rather than a single-sector focus.

  In this respect, our toolkit diverges from traditional precision livestock farming. We focus on welfare across species rather than production within a single sector, and place particular weight on positive welfare and positive affect, not only on the absence of suffering.

• Methods and approach

  Our methods combine ethograms and behavioural testing with cognitive assessment, stress and immune biomarkers (such as salivary cortisol), neurobiological and histological analysis, and machine-vision quantification under variable, real-world conditions. These data are integrated through statistical approaches including path, network and generalised linear mixed models.

  We build on established welfare-assessment frameworks and develop animal-based welfare indicators and protocols. Our approach is built to move methods from laboratory proof-of-concept, through practical test arenas like the The Livestock Production Research Centre (SHF), to deployed, validated tools that complement rather than replace professional human judgement.

• Research directions

  The group pursues four connected research directions.

  Our first direction focuses on establishing robust, operational ethograms and defining meaningful welfare indicators that capture both the presence of positive states and the absence of distress. These are grounded in the ethology and neurobiology of affect and cognition.

  Our second direction focuses on developing computer vision and machine learning models for reliable automated behaviour detection under real-world conditions, using confidence-rated outputs and cross-site validation.

  Our third direction focuses on translating continuous behavioural time series into actionable insight through trend analyses and early warning of health and welfare deviations, and practical decision support for farmers, veterinarians and animal keepers.

  Our fourth direction ensures that the resulting tools are designed with privacy, ethics, usability and human oversight at the forefront, with stakeholders involved directly in the co-design process.

• Collaboration

  Based at the Norwegian University of Life Sciences (NMBU) and working across animal science, veterinary medicine and technological departments, we connect academic researchers with research institutes, technology studios and industry partners.

  Our members contribute to national and international welfare standards and scientific-advisory work. We have led and participated in major European and Nordic research networks on environmental enrichment, stress, group housing and affective states across species, bringing high standing in the international applied-ethology community. This includes past leadership of its principal scientific society, as well as recognition through international research and innovation awards.

  By working at the intersection of behavioural and neural science, applied machine learning and real-world deployment, we aim to be a focal point for researchers, industry partners and advisors in policy and governance who use digital technology to raise animal-welfare standards, promote sustainable livestock production, and support lifelong learning across the sector.

We aim to be a focal point for researchers, industry partners and advisors in policy and governance who use digital technology to raise animal-welfare standards, promote sustainable livestock production. and support lifelong learning across the sector.

• Research-based teaching and methodology training

  The knowledge and technological methods we develop form the foundation of our teaching at NMBU. Several of our members are responsible for the mandatory animal welfare curriculum for veterinary students (VET353), helping ensure that future veterinarians build on an up-to-date, research-based foundation.

  In addition, we bring our computational expertise into doctoral training and lifelong learning. Our members are responsible for the PhD courses in research ethics and the practical use of artificial intelligence (VET400 and VET422), helping equip the next generation of researchers for responsible digital innovation.

• Members sorted alphabetically

• Associated Partners

  The group's associated partners contribute specialized expertise in quantitative genetics and artificial intelligence. This interdisciplinary collaboration is crucial to the group's goal of developing data-based, practical solutions that can be used to improve animal welfare in practice.

  Kristine Hov Martinsen is a researcher at Norsvin R&D, with expertise in quantitative genetics and pig breeding. She contributes to the group based on her experience using machine learning for the automatic detection of tail biting in pigs.

  Viko Murati is an AI-strategest and developer, founder of the Swiss AI studio fdk.ai. He combines advanced software development and artificial intelligence to create intuitive user interfaces, making complex data easily accessible to the end user.

  

  Kristine Hov Martinsen

  Researcher at Norsvin R&D

  • Quantitative Genetics
  • animal science
  • swine breeding
  • precision phenotyping

  

  Viko Murati

  AI Strategist & Developer

  • Applied AI
  • AI agents & automation
  • digital product development
  • AI strategy & consulting
  • deployment & operations

  Read more about Viko Murati's work at fdk.ai/en/

Our approach is built to move methods from laboratory proof-of-concept to deployed, validated tools that complement rather than replace professional human judgement. This ensures that digital innovation in welfare reaches the people who will actually use it, rather than stalling along the way.

Projects

• Selected recent scientific articles

  Rakun, J., Lepej, P., & Ocepek, M. (2026). Continuous RGB-D tracking enables assessment of activity, space use, and group synchrony in group-housed grower pigs under commercial conditions. Smart Agricultural Technology, 14, 102314.

  Ocepek, M. (2026). Within-hour incremental movement data of finishing pigs over 20 days using RFID–LoRaWAN technology. Scientific Data.

  Ocepek, M. (2026). Diurnal and day-to-day movement patterns of finishing pigs on deep straw bedding during the last 20 d before slaughter. Journal of Animal Science, 104, skag122.

  Lewis, R., Kostermans, T., Brovold, J. W., Laique, T., & Ocepek, M. (2025). Automated body condition scoring in dairy cows using 2D imaging and deep learning. AgriEngineering, 7(7), 241.

  Ocepek, M., Žnidar, A., Lavrič, M., Škorjanc, D., & Andersen, I. L. (2022). DigiPig: First developments of an automated monitoring system for body, head and tail detection in intensive pig farming. Agriculture, 12(1), 2.

Selected recent reviews and reports