Facial expression based automatic pain assessment in equines

This project focuses on data augmentation methods to improve facial expression-based automatic pain assessment in equines to alleviate the small data problem for the severe pain class. The data augmentation methods we explore range from image transformations, such as adding noise and cropping, to 3D mesh generation from a single image. We evaluate these methods and their effects on each part of the pain assessment pipeline. The pipeline consists of pose estimation, landmark detection and pain estimation.

Project leader

Jens Ruhof, MSc student

Academic supervisors

Prof. dr. A.A. (Albert) Salah, Prof. dr. R.C. (Remco) Veltkamp

Publications

Ruhof, J., A.A. Salah, T.J.P.A.M. van Loon, "Automatic pain estimation in equine faces: More effective uses for regions of interest," Proc. ACII Workshops, Glasgow, 2024