Self-supervised semantic segmentation is a crucial approach for addressing the issue of insufficient labeled data. However, traditional self-supervised learning methods designed for natural images are ...