Change Representation

Change representation in remote sensing and time series analysis focuses on effectively identifying and characterizing changes between different points in time. Current research emphasizes developing advanced deep learning models, including transformers and diffusion probabilistic models, often incorporating techniques like contrastive learning and incorporating visual-language models for improved semi-supervised learning. These advancements aim to improve the accuracy and efficiency of change detection, particularly in high-resolution imagery and complex scenarios, with applications ranging from disaster management to environmental monitoring. The field is actively exploring ways to leverage both global and local contextual information for more robust and detailed change maps.