Dynamic and Adaptive Self-Training for Semi-Supervised Remote Sensing Image Semantic Segmentation

Remote sensing (RS) technology has made remarkable progress, providing a wealth of data for various applications, such as ecological conservation and urban planning. However, the meticulous annotation of this data is labor-intensive, leading to a shortage of labeled data, particularly in tasks like semantic segmentation. Semi-supervised methods, combining consistency regularization (CR) with self-training, offer a solution to efficiently utilize labeled and unlabeled data. However, these methods encounter challenges due to imbalanced data ratios. To tackle these challenges, we introduce a self-training approach named dynamic and adaptive self-training (DAST), which is combined with dynamic pseudo-label sampling (DPS), distribution matching (DM), and adaptive threshold updating (ATU). DPS is tailored to address the issue of class distribution imbalance by giving priority to classes with fewer samples. Meanwhile, DM and ATU aim to reduce distribution disparities by adjusting model predictions across augmented images within the framework of CR, ensuring they align with the actual data distribution. Experimental results on the Potsdam and iSAID datasets demonstrate that DAST effectively balances class distribution, aligns model predictions with data distribution, and stabilizes pseudo-labels, leading to state-of-the-art performance on both datasets. These findings highlight the potential of DAST in overcoming the challenges associated with significant disparities in labeled-to-unlabeled data ratios.