.png){kind=icon}
返回
Hypergraph attentional convolutional neural network for salient object detection
.png){kind=icon}
PRE
AI
DOI：10.1007/s00371-022-02499-x.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Learning discriminative features and mining salient visual patterns play an important role in salient object detection (SOD) task. Existing SOD methods suffer from limited receptive field and insufficient cross-level feature mining. To this end, we propose a hypergraph attentional convolutional neural network for SOD task. Specifically, our method consists of (1) an attention based feature fusion module, which efficiently fuses lower layer as well as higher layer features, (2) a hypergraph-based long-range dependency encoder, which enhances the receptive field and global context for detection model, (3) a feature refinement layer, which highlights discriminative features and fuses attentional inputs, and (4) a dual iterative feature propagation decoder, which propagates features and upscale lower level feature maps to higher resolution. Both qualitative and quantitively experiments on public datasets verify the effectiveness of our proposed method. Compared with previous works, our model plays favorably against the state-of-the-arts methods.
Keyword:
Salient object detection
Cross-level feature mining
Hypergraph encoder
Feature propagation
期刊
IF:
2.9
论文数:
4.6K
被引数:
6.5K
机构
引用论文
.png){kind=icon}
LID, LEED, and Alternative Rating Systems—Integrating Low Impact Development Techniques with Green Building DesignLID，LEED和替代评级系统-将低影响开发技术与绿色建筑设计相结合
PREAI
分享收藏
查看更多内容