We propose a curriculum learning captioning method to caption fetal ultrasound images by training a model to dynamically transition between two different modalities (image and text) as training ...