NGram-MoSE: Efficient Remote Sensing Super-Resolution via N-Gram Context and Mixture-of-Experts
Researchers have developed NGram-MoSE, a new Transformer architecture for efficient super-resolution of remote sensing imagery. This model addresses the trade-off between spatial resolution and acquisition frequency in remote sensing data. NGram-MoSE utilizes N-Gram Context Injection for better local consistency and a Mixture-of-Experts design for scalable capacity with reduced computational cost. AI
IMPACT Introduces a more efficient method for enhancing remote sensing imagery, potentially improving downstream applications in environmental monitoring and disaster management.