EDL methods are more robust than individual deep learning (DL) models and measure uncertainty directly by highlighting the disagreement between base models.