Automatic Pain Estimation in Equine Faces: More Effective Uses for Regions of Interest

Recognition of pain in equines is essential for their welfare. There are several tools, such as the Horse Grimace Scale, EquiFACS and EQUUS-ARFAP, developed for pain assessment in equines, and there are approaches to automate assessment, as training observers takes time, and disagreements between observers are common. In this work, we provide a system for pain assessment in equine faces based on the EQUUS-ARFAP scale. The proposed system consists of three steps, namely, automatic detection of the facial regions, automatic head orientation detection, and automatic pain detection for each facial region of interest separately. Our main contribution is a detailed analysis of the usage of regions of interest as the main representation of the assessment pipeline, instead of facial landmarks. We show improved pain classification on the publicly available UU Equine Pain Face Dataset and advance the state of the art in this problem.