Dynamic context modeling based lightweight high-resolution network for dense prediction

Baoquan Sun, Qun Li, Ziyi Zhang

Abstract: Highlights•We present a novel efficient architecture for dense prediction, called Dynamic-HRNet.•Dynamic-HRNet consists of two key components, including a DSC and an ACM.•Two lightweight blocks are designed as the basic building units of Dynamic-HRNet.•The proposed Network achieves significant superiority for dense prediction.