Researchers introduced RingMo, a foundation model framework to enhance the accuracy of remote sensing image interpretation.
Remote sensing images are essential in various fields such as classification and change detection. There are a few limitations such as a domain gap between natural and remote sensing scenes and the poor generalization capacity of remote sensing models. Hence, the researchers envision implementing RingMo, a remote sensing foundation model framework that can enhance the benefits of generative self-supervised learning for remote sensing images. This model improves the accuracy of remote sensing image interpretation as per the Aerospace Information Research Institute (AIR), Chinese Academy of Sciences (CAS).
RingMo supports large-scale datasets constructed by collecting 2 million remote sensing images from satellite and aerial platforms, involving multiple scenes and objects around the world. This remote sensing foundation model training method is developed for dense and small objects in complex remote sensing scenes.
The deep learning approaches have also been included to instantly develop remote sensing image interpretation. ImageNet pre-trained models were also included to process remote sensing data for specified tasks but these implementations failed to meet the desired result. RingMo is the first generative foundation model that is implemented with masked image modeling which can be used for cross-modal remote sensing data. The researchers envision implementing this model for 3D reconstruction, residential construction, transportation, and other fields.
Click here for the Published Research Paper