A Genealogy of Foundation Models in Remote Sensing

Foundation models are increasingly being applied in remote sensing for representation learning, often leveraging successful computer vision techniques with minimal modifications. This approach reflects a trend where methods from computer vision are adapted to process remotely sensed data, as noted in recent research. Despite this progress, the field remains in a developing stage, with ongoing exploration of various competing methods to determine the most effective ways to utilize remote sensing data. The current landscape suggests that while foundational techniques provide a useful starting point, there is no consensus yet on best practices, indicating an active area of research and experimentation. This evolving status highlights the dynamic nature of foundation model applications in remote sensing, as researchers continue to refine and optimize these approaches.