Transformer with large convolution kernel decoder network for salient object detection in optical remote sensing images

Pengwei Dong, Bo Wang, Runmin Cong, Hai-Han Sun, Chongyi Li

Abstract: Highlights•The proposed TLCKD-Net can model the long-range dependencies for feature extraction.•The Transformer backbone network and the decoding module with large convolutional kernels are constructed.•The multi-order gated refinement module is capable of refining features of salient objects.•The salient feature enhancement module can better distinguish refined features from background.