自适应频率泛锐化与专家混合
原文约400字，阅读约需1分钟。发表于：。Pan-sharpening research has not explored the potential solution in the frequency domain. This paper proposes the Frequency Adaptive Mixture of Experts (FAME) learning framework, which effectively reconstructs low-frequency and high-frequency information and adapts to remote sensing images with content variations, outperforming other state-of-the-art methods.
本文介绍了一种频率感知的掩码自编码器（bioFAME），用于全面建模多模式生物信号。bioFAME在预训练过程中充分利用多模态信息，适应不同任务和模态。实验结果显示，与之前方法相比，bioFAME在分类准确度上平均提升了5.5％，且具有模态不匹配的稳健性。
