Reverse Attention for Salient Object Detection

By Shuhan Chen, Xiuli Tan, Benge Wang, Xuelong Hu. European Conference on Computer Vision, 2018

641 citations recorded by the source corpus.

Benefit from the quick development of deep learning techniques, salient object detection has achieved remarkable progresses recently. However, there still exists following two major challenges that hinder its application in embedded devices, low resolution output and heavy model weight. To this end, this paper presents an accurate yet compact deep network for efficient salient object detection. More specifically, given a coarse saliency prediction in the deepest layer, we first employ residual learning to learn side-output residual features for saliency refinement, which can be achieved with very limited convolutional parameters while keep accuracy. Secondly, we further propose reverse attention to guide such side-output residual learning in a top-down manner. By erasing the current predicted salient regions from side-output features, the network can eventually explore the missing object parts and details which results in high resolution and accuracy. Experiments on six benchmark datasets demonstrate that the proposed approach compares favorably against state-of-the-art methods, and with advantages in terms of simplicity, efficiency (45 FPS) and model size (81 MB).

Subfields: Computer Vision and Pattern Recognition, Human-Computer Interaction

View Paper: https://www.semanticscholar.org/paper/ba23953a294108c05b1cecd023156b1b7f994b22

How researchers see this paper

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

In Computer Vision and Pattern Recognition

In Human-Computer Interaction

Citations in the corpus

CaraNet: context axial reverse attention network for segmentation of small medical objects (2021, methodology): “Therefore, we progressively mine discriminative tissue region through an erasure of the foreground object by applying reverse attention [8].”
Recent advances and clinical applications of deep learning in medical image analysis (2021, methodology): “Aside from the semi-supervised learning strategy, there are three critical components in the model responsible for the good performance: parallel partial decoder (PPD) (Wu et al., 2019a), reverse attention (RA) (Chen et al., 2018b), and edge attention (Zhang et al., 2019).”
Multi-Scale Interactive Network for Salient Object Detection (2020, methodology): “…methods, including the SRM [32], PiCANet [20], DGRL [34], R3Net [6], CapSal [50], BAS-Net [27], BANet [30], ICNet [36], CPD [41], Amulet [51], NLDF [23], MSRNet [15], UCF [52], DSS [13], PAGR[53], RAS [4], C2SNet [17], HRS [46], PAGE [37], MLM-SNet [40], AFNet [9], SCRN [42], and EGNet [56].”
Suppress and Balance: A Simple Gated Network for Salient Object Detection (2020, methodology): “We compare the proposed algorithm with seventeen state-of-the-art saliency detection methods, including the DCL [26], DSS [20], Amulet [69], SRM [51], DGRL [53], RAS [7], PAGRN [70], BMPM [67], R3Net [12], HRS [66], MLMS [58], PAGE [57], ICNet [55], CPD [59], BANet [45], BASNet [37] and Capsal [68].”
Camouflaged Object Segmentation with Distraction Mining (2021, methodology): “Distraction cues have been explored in many vision tasks such as salient object detection [4, 56], semantic segmentation [21] and visual tracking [73].”
Medical Image Segmentation via Cascaded Attention Decoding (2023, methodology): “[10] adopt reverse attention [6] for polyp segmentation.”
Pattern Recognition (2018, neutral): “Soft biometrics can improve the recognition accuracy of a biometric system by fusing with primary biometric traits [2] .”
MISSFormer: An Effective Medical Image Segmentation Transformer (2021, methodology): “Besides, various attention mechanisms were effective to promote segmentation performance, reverse attention(Chen et al. 2018) was applied to accurate polyp segmentation(Fan et al.”
Inf-Net: Automatic COVID-19 Lung Infection Segmentation From CT Images (2020, methodology): “Second, we propose a progressive framework, acting as the fine labeler, to mine discriminative infection regions in an erasing manner [53], [54].”
Interactive Two-Stream Decoder for Accurate and Fast Saliency Detection (2020, methodology): “We compare our model with 16 existing saliency detection models, including 10 nearly real-time models, such as RFCN [32], Amulet [45], UCF [46], NLDF [24], CKT [18], BMP [44], PCA [22], PAGE [35], DSS [12], EGNet [48] and 6 real-time models, such as RA [3], AFNet [8], BASNet [26], CPD [37], PoolNet ”
Stacked Cross Refinement Network for Edge-Aware Salient Object Detection (2019, methodology): “Liu et al. [22] [4] propose a reverse attention network, which eventually explores the missing object parts and details by erasing the current predicted salient regions from side-output features.”
Attentive Feedback Network for Boundary-Aware Salient Object Detection (2019, methodology): “Ours achieves the real-time speed of 26 FPS which is comparable with RAS (33 FPS) and DHS (28 FPS).”
EGNet: Edge Guidance Network for Salient Object Detection (2019, methodology): “j∈Y− logPr(yj = 0|Ĝ ;W (i) D ), i ∈ [3, 6].”
F3Net: Fusion, Feedback and Focus for Salient Object Detection (2019, methodology): “Many CNNs-based models (Hou et al. 2019; Qin et al. 2019; Chen et al. 2018; Wu, Su, and Huang 2019; Li et al. 2018; Zhang et al. 2018b; Feng, Lu, and Ding 2019; Wang et al. 2017b; Zhang et al. 2017) have achieved remarkable progress and pushed the performance of SOD to a new level.”
Multi-Scale Self-Guided Attention for Medical Image Segmentation (2019, background): “tion [14], object detection [15] and classification [16].”