LF3Net: Leader-follower feature fusing network for fast saliency detection

doi:10.1016/j.neucom.2021.03.080

CORC > 长春光学精密机械与物理研究所 > 中国科学院长春光学精密机械与物理研究所

	LF3Net: Leader-follower feature fusing network for fast saliency detection
	H. Luo; G. Han; X. Wu; P. Liu; H. Yang and X. Zhang
刊名	Neurocomputing
	2021
卷号	449 页码:24-37
ISSN号	9252312
DOI	10.1016/j.neucom.2021.03.080
英文摘要	Recently, convolutional neural networks (CNNs) have been widely used for saliency detection. Most of existing saliency detection methods produce saliency maps from the complementary multi-level convolutional features. However, it is still a challenging task to accurately integrate multi-level features for saliency detection. In this paper, we explore the intrinsic relationships between multi-level features and introduce the Stackelberg game theory as a new strategy to fuse multi-level features for saliency detection. Based on the theory, we propose a leader-follower feature fusing network (LF3Net) to obtain saliency maps. We first apply a multi-scale context-aware leader-follower attention module (MCLAM) to select multi-scale spatial and semantic information. Then, we propose a leader-follower feature fusing module (LF3M) to integrate the multi-level features. Extensive experiments on five datasets show that the proposed method outperforms the state-of-the-art approaches under different evaluation metrics. In addition, our network can run fast at the real-time speed of 75 FPS. 2021 Elsevier B.V.
URL标识	查看原文
内容类型	期刊论文
源URL	[http://ir.ciomp.ac.cn/handle/181722/65376]
专题	中国科学院长春光学精密机械与物理研究所
推荐引用方式 GB/T 7714	H. Luo,G. Han,X. Wu,et al. LF3Net: Leader-follower feature fusing network for fast saliency detection[J]. Neurocomputing,2021,449:24-37.
APA	H. Luo,G. Han,X. Wu,P. Liu,&H. Yang and X. Zhang.(2021).LF3Net: Leader-follower feature fusing network for fast saliency detection.Neurocomputing,449,24-37.
MLA	H. Luo,et al."LF3Net: Leader-follower feature fusing network for fast saliency detection".Neurocomputing 449(2021):24-37.