Accurate salient object detection via dense recurrent connections and residual-based hierarchical feature integration

Yanpeng Cao, Guizhong Fu, Jiangxin Yang, Yanlong Cao, Michael Ying Yang

Research output: Contribution to journalArticleAcademicpeer-review

Abstract

Recently, the convolutional neural network (CNN) has achieved great progress in many computer vision tasks including object detection, image restoration, and scene understanding. In this paper, we propose a novel CNN-based saliency detection method through dense recurrent connections and residual-based hierarchical feature integration. Inspired by the recent neurobiological finding that abundant recurrent connections exist in the human visual system, we firstly propose a novel dense recurrent CNN module (D-RCNN) to learn informative saliency cues by incorporating dense recurrent connections into sub-layers of convolutional stages. Then we present a residual-based architecture with short connections for deep supervision which hierarchically combines both coarse-level and fine-level feature representations. Our end-to-end method takes raw RGB images as input and directly outputs saliency maps without relying on any time-consuming pre/post-processing techniques. Extensive qualitative and quantitative evaluation results on four widely tested benchmark datasets demonstrate that our method can achieve more accurate saliency detection results solutions with significantly fewer model parameters.

Original languageEnglish
Pages (from-to)103-112
Number of pages10
JournalSignal Processing: Image Communication
Volume78
DOIs
Publication statusPublished - 1 Oct 2019

Fingerprint

Neural networks
Image reconstruction
Computer vision
Processing
Object detection

Keywords

  • Convolutional neural network
  • Deep supervision
  • Hierarchical feature fusion
  • Recurrent convolutional layer
  • Salient object detection
  • ITC-ISI-JOURNAL-ARTICLE

Cite this

@article{37fba537a9fd4e1d89413edd29d3dd24,
title = "Accurate salient object detection via dense recurrent connections and residual-based hierarchical feature integration",
abstract = "Recently, the convolutional neural network (CNN) has achieved great progress in many computer vision tasks including object detection, image restoration, and scene understanding. In this paper, we propose a novel CNN-based saliency detection method through dense recurrent connections and residual-based hierarchical feature integration. Inspired by the recent neurobiological finding that abundant recurrent connections exist in the human visual system, we firstly propose a novel dense recurrent CNN module (D-RCNN) to learn informative saliency cues by incorporating dense recurrent connections into sub-layers of convolutional stages. Then we present a residual-based architecture with short connections for deep supervision which hierarchically combines both coarse-level and fine-level feature representations. Our end-to-end method takes raw RGB images as input and directly outputs saliency maps without relying on any time-consuming pre/post-processing techniques. Extensive qualitative and quantitative evaluation results on four widely tested benchmark datasets demonstrate that our method can achieve more accurate saliency detection results solutions with significantly fewer model parameters.",
keywords = "Convolutional neural network, Deep supervision, Hierarchical feature fusion, Recurrent convolutional layer, Salient object detection, ITC-ISI-JOURNAL-ARTICLE",
author = "Yanpeng Cao and Guizhong Fu and Jiangxin Yang and Yanlong Cao and Yang, {Michael Ying}",
year = "2019",
month = "10",
day = "1",
doi = "10.1016/j.image.2019.06.004",
language = "English",
volume = "78",
pages = "103--112",
journal = "Signal Processing: Image Communication",
issn = "0923-5965",
publisher = "Elsevier",

}

Accurate salient object detection via dense recurrent connections and residual-based hierarchical feature integration. / Cao, Yanpeng; Fu, Guizhong; Yang, Jiangxin; Cao, Yanlong; Yang, Michael Ying.

In: Signal Processing: Image Communication, Vol. 78, 01.10.2019, p. 103-112.

Research output: Contribution to journalArticleAcademicpeer-review

TY - JOUR

T1 - Accurate salient object detection via dense recurrent connections and residual-based hierarchical feature integration

AU - Cao, Yanpeng

AU - Fu, Guizhong

AU - Yang, Jiangxin

AU - Cao, Yanlong

AU - Yang, Michael Ying

PY - 2019/10/1

Y1 - 2019/10/1

N2 - Recently, the convolutional neural network (CNN) has achieved great progress in many computer vision tasks including object detection, image restoration, and scene understanding. In this paper, we propose a novel CNN-based saliency detection method through dense recurrent connections and residual-based hierarchical feature integration. Inspired by the recent neurobiological finding that abundant recurrent connections exist in the human visual system, we firstly propose a novel dense recurrent CNN module (D-RCNN) to learn informative saliency cues by incorporating dense recurrent connections into sub-layers of convolutional stages. Then we present a residual-based architecture with short connections for deep supervision which hierarchically combines both coarse-level and fine-level feature representations. Our end-to-end method takes raw RGB images as input and directly outputs saliency maps without relying on any time-consuming pre/post-processing techniques. Extensive qualitative and quantitative evaluation results on four widely tested benchmark datasets demonstrate that our method can achieve more accurate saliency detection results solutions with significantly fewer model parameters.

AB - Recently, the convolutional neural network (CNN) has achieved great progress in many computer vision tasks including object detection, image restoration, and scene understanding. In this paper, we propose a novel CNN-based saliency detection method through dense recurrent connections and residual-based hierarchical feature integration. Inspired by the recent neurobiological finding that abundant recurrent connections exist in the human visual system, we firstly propose a novel dense recurrent CNN module (D-RCNN) to learn informative saliency cues by incorporating dense recurrent connections into sub-layers of convolutional stages. Then we present a residual-based architecture with short connections for deep supervision which hierarchically combines both coarse-level and fine-level feature representations. Our end-to-end method takes raw RGB images as input and directly outputs saliency maps without relying on any time-consuming pre/post-processing techniques. Extensive qualitative and quantitative evaluation results on four widely tested benchmark datasets demonstrate that our method can achieve more accurate saliency detection results solutions with significantly fewer model parameters.

KW - Convolutional neural network

KW - Deep supervision

KW - Hierarchical feature fusion

KW - Recurrent convolutional layer

KW - Salient object detection

KW - ITC-ISI-JOURNAL-ARTICLE

UR - http://www.scopus.com/inward/record.url?scp=85067870516&partnerID=8YFLogxK

UR - https://ezproxy2.utwente.nl/login?url=https://library.itc.utwente.nl/login/2019/isi/michaelyang_acc.pdf

U2 - 10.1016/j.image.2019.06.004

DO - 10.1016/j.image.2019.06.004

M3 - Article

VL - 78

SP - 103

EP - 112

JO - Signal Processing: Image Communication

JF - Signal Processing: Image Communication

SN - 0923-5965

ER -