BMANet: Boundary-guided multi-level attention network for polyp segmentation in colonoscopy images

Zihuang Wu, Hua Chen, Xinyu Xiong, Shang Wu, Hongwei Li, Xinyu Zhou

Abstract: Highlights•We propose a novel boundary-guided segmentation network for polyp segmentation.•A BAM module is proposed to aggregate low-level and high-level features.•A BMA module is proposed to better process polyp region and boundary.•Extensive experiments demonstrate that our BMANet achieves SOTA performance.