Deep learning models based on diffusion processes have shown great potential in a range of generative tasks, such as image generation. For remote sensing applications, generative models are not that common. The question that we tried to answer is whether diffusion processes can be used to efficiently pretrain models for discriminative tasks?

Multimodal datasets for remote sensing are oftentimes limited to two data modalities, such as multispectral and SAR polarization data. In order to experiment with a much wider range of data modalities, we extended the well-known BigEarthNet dataset to includes a wide range of data modalities.

Self-supervised learning provides a powerful means to pretrain models based on un-labeled data. Un-labeled Earth observation data are abundant: this circumstance combined with the availability of multi modal data makes Earth observation a perfect playground for self-supervised learning. Our early results are very promising…