Remote Sensing Image Change Captioning

The field studies how to compare images of the same place taken at different times and generate captions that identify what changed and where. Common approaches model temporal differences with attention, multiscale feature fusion, directional representations of appearance and disappearance, or structured detection cues; some adapt vision-language models and large language models with task-specific training. Methods are evaluated on benchmarks such as LEVIR-CC, SECOND-CC, and WHU-CDC using caption-quality metrics, with newer datasets adding change masks and question answering. Key challenges include distinguishing real change from viewpoint or scale differences, producing factually grounded descriptions, and balancing accuracy with efficient deployment.