SCL: Towards Domain Generalization via Single-Temporal Multimodal Contrastive Learning for Remote Sensing Change Detection
Researchers have developed a new method called SCL for change detection in remote sensing imagery, aiming to improve cross-dataset generalization. This approach utilizes a single-temporal multimodal contrastive learning strategy, leveraging visual-language pre-training models. SCL addresses the need for large amounts of paired labeled data by training on single-temporal images without requiring target dataset-specific training, demonstrating superior performance over existing methods. AI
IMPACT Enhances generalization for remote sensing change detection, potentially reducing data labeling costs.