Project

Project summary

Our project aims to develop technology for automated assessment of animal emotional wellbeing through the interpretation of body language, informed by work in human activity and emotion recognition using AI. Body posture and movement, which together characterise a wider body language, mani-fest as varying body shapes/poses and movements. We will utilise human eye tracking to identify holistic spatial temporal animal features linked to perceived animal emotional states, then apply this knowledge via a novel form of supervised machine learning.

Coming together in close collaboration across WPs our interdisciplinary team will explore computer vision-based machine learning techniques to train machine models with numerous examples of pig body language corresponding to different emotional states (ground truth), such as positive emotions associated with reward or negative emotions in unfamiliar environments. With extensive examples including structured treatments and environments to create a spectrum of emotional experiences, we aim to correlate specific stances and movements with given emotions and levels of arousal.

Automated appraisal of affective states (short-term emotions and long-term moods) offers opportunity for continuous welfare monitoring, overcoming human time limitations and enabling 24/7 monitoring of numerous individual animals. This system could integrate with existing automated assessments (eg for weight, body condition, lameness) implemented by our team[12]. It would complement welfare monitoring by humans and facilitate algorithmic learning about individual animals, tracking their trajectories as their experiences and environmental interactions evolve over time. Through a system of detected trends and alerts, intelligent automation could offer a powerful means to identify animals needing follow-up interventions in a timely manner, ensuring animal wellbeing, reduc-ing production costs (eg treatments, morbidity) and antimicrobial usage (eg fewer blanket treatments).