SAR Target Detection
DenoDet: Attention as Deformable Multi-Subspace Feature Denoising for Target Detection in SAR Images
This study presents DenoDet, a novel synthetic aperture radar (SAR) target detection network that leverages frequency domain transform for multi-scale subspace representation. The proposed TransDeno module dynamically denoises across subspaces by preserving target signals and attenuating noise, and a deformable group fully-connected layer (DeGroFC) adjusts the granularity of subspace processing based on input features.
