Accurately assessing animal emotion remains difficult because emotional cues vary across species, lack standardized metrics, and are prone to human observer bias. This paper synthesizes multidisciplinary insights from an international workshop evaluating how artificial intelligence and computer vision can measure non-invasive visual behavior. Focusing on species such as dogs, non-human primates, horses, and livestock, the study examines anatomical frameworks like Facial Action Coding Systems alongside deep learning pose estimation tools. The study highlights how automated tools detect subtle facial and bodily indicators while offering essential strategies to mitigate biases in human observation and AI training data. Overall, this work provides a foundational framework for developing objective, reliable emotion recognition technologies in animal welfare and behavior research.
Prototype-based Explainable Deep Learning for Sex Classification of Mountain Gazelles in the Wild
Standard deep learning tools in automated wildlife monitoring often function
