.png){kind=icon}
返回
Interactive segmentation based on multiscale feature cascading
.png){kind=icon}
PRE
AI
DOI：10.1007/s10489-024-05824-0.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
In this paper, we explore a principal method to enhance image segmentation quality through limited user interaction. We propose a model solution called the multiscale feature cascading network (MFC-Net), which effectively leverages annotated information and enhances segmentation performance in complex scenes. First, we convert the user-provided click information into a disk map, using two different disk radii to capture interaction influences within different ranges. Then, we employ a dual-channel attention module via multiscale feature cascading. Finally, we devise a refinement module to improve the segmentation results. We validated the effectiveness of MFC-Net on four commonly used image segmentation datasets. Extensive experiments show that MFC-Net could better perceive user's intentions and significantly reduce the burden of user interaction.
Keyword:
Interactive segmentation
Multiscale feature cascading
Channel attention block
Refinement module
期刊
IF:
3.5
论文数:
7.6K
被引数:
1.7W
机构
引用论文
A fully convolutional two-stream fusion network for interactive image segmentation.png)
NEURAL NETWORKS
OAAI
NEURAL NETWORKSIF6.3
OAAI分享收藏Distant regulatory elements in a Sox10‐βGEO BAC transgene are required for expression of Sox10 in the enteric nervous system and other neural crest‐derived tissuesSox10-βgeo BAC转基因中的远距离调控元件是肠神经系统和其他神经源性组织中 Sox10 表达所必需的OAAI
OAAI分享收藏
分享收藏
查看更多内容