Abstract: In computer vision, neural network models typically require a large amount of manually annotated images or video data for training. To reduce annotation costs, self-supervised learning has ...