Beyond Pixel-Level Annotation: Exploring Self-Supervised Learning for Change Detection With Image-Level Supervision

Change detection (CD) in high-resolution remote sensing has received large attention due to its wide range of applications. Many methods have been proposed in the literature and achieved excellent performance. However, they are often fully supervised, thus requiring abundant pixel-level labeled samples, which is time-consuming and labor-intensive. Especially compared to the common single-temporal interpretation, labeling bi-temporal images is often more complicated. Therefore, this study combines weakly supervised learning (WSL) to reduce label acquisition costs. However, changed regions are small, fragmented, and similar to the background, which increases the gap between weakly supervised and fully supervised tasks. To address these difficulties, we explore self-supervised methods to construct a WSL framework based on image-level labels for general CD, termed WSLCD in this article. First, we design a double-branch Siamese network to derive embeddings and initial class attention maps (...

Change detection (CD) in high-resolution remote sensing has received large attention due to its wide range of applications. Many methods have been proposed in the literature and achieved excellent performance. However, they are often fully supervised, thus requiring abundant pixel-level labeled samples, which is time-consuming and labor-intensive. Especially compared to the common single-temporal interpretation, labeling bi-temporal images is often more complicated. Therefore, this study combines weakly supervised learning (WSL) to reduce label acquisition costs. However, changed regions are small, fragmented, and similar to the background, which increases the gap between weakly supervised and fully supervised tasks. To address these difficulties, we explore self-supervised methods to construct a WSL framework based on image-level labels for general CD, termed WSLCD in this article. First, we design a double-branch Siamese network to derive embeddings and initial class attention maps (CAMs), which input the original image pair and the spatially transformed image pair. Second, mutual learning and equivariant regularization (MLER) are enforced on CAMs from different views, which implements consistency constraints in confusion regions and makes CAMs learn from each other based on saliency regions. Furthermore, prototype-based contrastive learning (PCL) is designed such that unreliable pixels can learn from prototypes computed from reliable pixels. PCL includes intraview contrast and cross-view contrast depending on whether the prototypes and class embeddings are from the same view. With the above strategies, we narrow the gap between image-level weakly supervised CD and fully supervised CD. Experiments are conducted on three CD datasets, including CLCD, DSIFN, and GCD. Our method achieves state-of-the-art performance on pseudo-label generation and CD. The code is available at https://github.com/mfzhao1998/WSLCD.