Researchers have introduced MetaEarth-MM, a novel generative model designed for unified multimodal remote sensing image generation. This model addresses the scarcity of complete paired observations by enabling joint generation and any-to-any translation across five modalities within a single framework. MetaEarth-MM operates by first inferring a latent scene representation and then generating target modalities based on this representation, moving beyond direct appearance-level cross-modal mapping. To facilitate training, a large-scale dataset named EarthMM, containing 2.8 million multi-resolution global images, has also been constructed. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Enables more comprehensive analysis of Earth observation data by unifying multiple remote sensing modalities.
RANK_REASON The cluster contains a new academic paper detailing a novel model and dataset. [lever_c_demoted from research: ic=1 ai=1.0]