MDCA-Net: Remote Sensing Image Classification Based on Multiscale Dynamic Convolution and Attention Enhancement (opens in new tab)

In response to the challenges faced by traditional deep learning (DL) models in adaptively extracting multiscale features from remote sensing images and establishing long-range dependencies across both channels and spatial dimensions—issues that contribute to a decline in classification performance—this study introduces a novel network architecture that integrates dynamic multiscale attention with spatiotemporal fusion, designated as multiscale dynamic convolution and attention fusion network...