DC-Net: Divide-and-conquer for salient object detection

Jiayi Zhu, Xuebin Qin, Abdulmotaleb Elsaddik

Abstract: Highlights•Divide-and-Conquer for salient object detection.•Parallel versions of ResNet and Swin-Transformer are implemented by reparameterization.•A new multi-scale contextual module improves the disadvantage of ASPP and U-shape structure.•Edge maps with different edge widths can impact the model performance.•The pipeline for high-resolution and high-quality segmentation for the meticulous fields of medical, aviation, and military.