Deep Contrast Learning for Salient Object Detection

By Guanbin Li, Yizhou Yu. Computer Vision and Pattern Recognition, 2016

768 citations recorded by the source corpus.

Salient object detection has recently witnessed substantial progress due to powerful features extracted using deep convolutional neural networks (CNNs). However, existing CNN-based methods operate at the patch level instead of the pixel level. Resulting saliency maps are typically blurry, especially near the boundary of salient objects. Furthermore, image patches are treated as independent samples even when they are overlapping, giving rise to significant redundancy in computation and storage. In this paper, we propose an end-to-end deep contrast network to overcome the aforementioned limitations. Our deep network consists of two complementary components, a pixel-level fully convolutional stream and a segment-wise spatial pooling stream. The first stream directly produces a saliency map with pixel-level accuracy from an input image. The second stream extracts segment-wise features very efficiently, and better models saliency discontinuities along object boundaries. Finally, a fully connected CRF model can be optionally incorporated to improve spatial coherence and contour localization in the fused result from these two streams. Experimental results demonstrate that our deep model significantly improves the state of the art.

Subfields: Cognitive Neuroscience, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/207f85142c44efe3ef0bb13a12f3cc2973b68801

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Cognitive Neuroscience

In Computer Vision and Pattern Recognition

Citations in the corpus

A Review on Deep Learning Techniques Applied to Semantic Segmentation (2017, methodology): “The RGB pipeline relies on a variant of the DeepLab architecture [29] concatenating features at three different scales to enrich feature representation (inspired by [100]).”
UberNet: Training a Universal Convolutional Neural Network for Low-, Mid-, and High-Level Vision Using Diverse Datasets and Limited Memory (2016, methodology): “Additional datasets are typically used to benchmark this task, e.g. in [58], we will explore the performance of our method on those datasets in the future.”
A Simple Pooling-Based Design for Real-Time Salient Object Detection (2019, methodology): “Image GT Ours PiCANet[24] DGRL[38] PAGR[46] SRM[37] Amulet[44] DSS[9] MSR[17] DCL[19]”
Reverse Attention for Salient Object Detection (2018, methodology): “In addition, some time consuming post-processing skills are also applied for refinement, e.g., superpixel-based filter [10], fully connected conditional random field (CRF) [22,8,17].”
BBS-Net: RGB-D Salient Object Detection with a Bifurcated Backbone Strategy Network (2020, methodology): “tps://github.com/zyjwuyan/BBS-Net. 2 RELATED WORKS 2.1 Salient Object Detection Over the past several decades, SOD [38]–[40] has garnered signiﬁcant research attention due to its diverse applications [41]–[43]. In early years, SOD methods were mainly based on intrinsic prior knowledge such as center”
Learning to Detect Salient Objects with Image-Level Supervision (2017, methodology): “Unsupervised Weakly Fully MR wCtr MBS BSCA WSS LEGS MDF MC DS SELD DCL RFCN ECSSD 0.690 0.676 0.673 0.705 0.823 0.785 0.807 0.796 0.826 0.810 0.829 0.834 SED 0.782 0.786 0.776 0.756 0.838 0.800 0.795 0.817 0.794 0.815 0.825 0.813 PASAL-S 0.583 0.597 0.604 0.597 0.720 – 0.705 0.687 0.655 0.714 0.710 ”
RGBD Salient Object Detection via Deep Fusion (2016, methodology): “As in the proposed method, CNNs have been adopted in several other works to extract hierarchical feature representations for detecting salient regions [38], [39], [40], [41], [42], [43].”
Pyramid Dilated Deeper ConvLSTM for Video Salient Object Detection (2018, methodology): “, multi-stream network with embedded superpixels [22, 25], recurrent module [26, 40], and multi-scale and hierarchical feature fusion [16, 51, 36], etc.”
Exploiting Saliency for Object Segmentation from Image Level Labels (2017, methodology): “Like most computer vision areas, hand-crafted methods [14, 28, 8] have now been surpassed by convnet based approaches [53, 22, 21] for object saliency.”
Nested Network With Two-Stream Pyramid for Salient Object Detection in Optical Remote Sensing Images (2019, background): “Li and Yu [23] proposed a deep contrast network for saliency detection, where the multi-scale fully convolutional stream captures the visual”
Enhanced-alignment Measure for Binary Foreground Map Evaluation (2018, methodology): “…including 3 traditional ones (ST [Liu et al. , 2014], DRFI [Wang et al. , 2017], and DSR [Li et al. , 2013]) and 7 deep learning based ones (DCL [Li and Yu, 2016], RFCN [Wang et al. , 2016], MC [Zhao et al. , 2015], MDF [Li and Yu, 2015], DISC [Chen et al. , 2016], DHS [Liu and Han, 2016], and…”
Advanced Deep-Learning Techniques for Salient and Category-Specific Object Detection: A Survey (2018, methodology): “We selected seven models that were published in 2015 and 2016 and have released their codes or computed saliency maps: LEGS [46], DHS [20], DCL [48], ELD [47], MCDL [87], MDF [83], and RFCN [49].”
RGB-D salient object detection: A survey (2020, background): “progress has been made in the SOD field over the past several years [30]–[36], [36]–[44], there is still room for improvement when faced with challenging factors, such as complicated background or different lighting conditions in the scenes.”
Attentive Feedback Network for Boundary-Aware Salient Object Detection (2019, result): “We compared our algorithm with other 12 state-ofthe-art deep learning methods, which are LEGS [26], RFCN [28], ELD [13], DCL [16], DS [18], DHS [21], Amulet [33], DSS [10], C2SNet [17], RAS [5], DGRL [29] and PiCANet [22].”
Pyramid Feature Attention Network for Saliency Detection (2019, methodology): “To be speciﬁc, (1) the proposed method not only highlights the correct salient object regions clearly, but also well suppresses the saliencies [42] 0.8508 0.0484 0.9249 0.0478 0.9200 0.0392 0.8806 0.0788 0.7740 0.0635 GRL[33] 0.8341 0.0509 0.9230 0.0446 0.9130 0.0377 0.8811 0.0799 0.7788 0.0632 PAGR”
Deeply Supervised Salient Object Detection with Short Connections (2016, methodology): “We adopt the same technique as in [35].”
Contrast Prior and Fluid Pyramid Integration for RGBD Salient Object Detection (2019, methodology): “The other one located at the lower right is the simple multi-scale fusion architecture [36] (M in Tab.”