New AI research to revolutionise animal welfare

An international project led by UWE Bristol's Centre for Machine Vision (CMV) will aim to fuse human behavioural science with AI, taking a step towards technology that understands not just what animals do, but how they feel.

The research programme named 'HoliWell: Towards a Holistic Assessment of Animal Welfare using Emotion and Deep Learning', teaches machines to interpret animal emotion in the same way an experienced observer might.

Funded through Defra in the UK via the European Partnership on Animal Health and Welfare, the project will see five European partners working together to develop human-like, holistic models that read animals' facial cues as well as posture, body language, and behaviour - recognising both momentary emotions and longer-term moods.

One way in which animals express their emotions is through how they move and carry themselves. The research team will use eye tracking to study where people look when watching animals to find patterns that relate to different emotions. Then, they'll train a computer to recognise these patterns using a new kind of machine learning.

The interdisciplinary team will combine insights from animal behaviour, psychology, and computer science to build training datasets that capture how pigs physically express themselves across a broad range of emotional situations. By designing controlled environments that ethically evoke different experiences, from rewarding to challenging, they will gather clear examples of body language linked to known emotional states. These diverse recordings will then help them match specific body pose and movement patterns to particular emotions and levels of arousal.

Once trained, the AI system will generate automated alerts and trend detection for individual animals - potentially helping identify those needing follow-up care earlier, reducing morbidity, lowering treatment costs, and contributing to more precise and compassionate welfare management.

This research builds on over a decade of pioneering work by CMV, which has become a global leader in using machine vision to enhance animal welfare.

Professor Melvyn Smith, co-director of the Centre for Machine Vision, said: "Every one of CMV's initiatives contributes to a future in which farmers can use intelligent systems to ensure that animals live healthier, more contented lives. Through the fusion of AI and ethology - the scientific study of animal behaviour - we're helping society move closer to truly understanding, and improving, the emotional lives of the animals in our care."

CMV began its research with the Precision Cow Health Management project (2013-2023), replacing manual checks with automated 3D imaging. This led to HerdVision, a contact-free system for estimating weight, monitoring cow condition and mobility, now used across Europe, the US, New Zealand, and Uruguay. Validated by Arla UK 360, HerdVision won the 2023 Royal Dairy Innovation Award and received an Outstanding rating from the Knowledge Transfer Partnership Panel.

Next came a project exploring the use of deep learning to detect early signs of lameness in dairy cattle caused by digital dermatitis (2022-2024). This technology powers PediVue, a commercial system helping UK and European farmers prevent mobility issues. PediVue earned the 2025 Royal Dairy Innovation Award for advancing proactive welfare.

CMV has also focused on pigs, undertaking EmotiPig (2018-2021) to assess emotional states via facial expression recognition and deep learning. This led to FarmCare (2022-2025), which supports antimicrobial reduction through detection of animal stress, and an involvement in Pig-ID (2023-2025), enabling long-term biometric tracking.

Ongoing research also includes IntelliPig (2023-2026), which combines facial recognition, emotion detection, and body condition analysis into a single intelligent monitoring station. Funded by Innovate UK, IntelliPig aims to continuously analyse each pig's face and body to monitor health and infer emotional wellbeing - offering farmers individualised welfare profiles and early alerts.

"Through our ongoing work, we have demonstrated the power of cross-disciplinary collaboration, uniting computer vision engineers, animal behaviourists, data scientists, and commercial partners to deliver innovation with real-world impact," added Professor Smith. "Our projects span the spectrum from applied product development to fundamental science, and all share a single vision: to make animal welfare measurable, meaningful, and manageable."