SuperCNN: A Superpixelwise Convolutional Neural Network for Salient Object Detection

By Shengfeng He, Rynson W. H. Lau, Wenxi Liu, Zhe Huang, Qingxiong Yang. 2015

297 citations recorded by the source corpus.

Existing computational models for salient object detection primarily rely on hand-crafted features, which are only able to capture low-level contrast information. In this paper, we learn the hierarchical contrast features by formulating salient object detection as a binary labeling problem using deep learning techniques. A novel superpixelwise convolutional neural network approach, called SuperCNN, is proposed to learn the internal representations of saliency in an efficient manner. In contrast to the classical convolutional networks, SuperCNN has four main properties. First, the proposed method is able to learn the hierarchical contrast features, as it is fed by two meaningful superpixel sequences, which is much more effective for detecting salient regions than feeding raw image pixels. Second, as SuperCNN recovers the contextual information among superpixels, it enables large context to be involved in the analysis efficiently. Third, benefiting from the superpixelwise mechanism, the required number of predictions for a densely labeled map is hugely reduced. Fourth, saliency can be detected independent of region size by utilizing a multiscale network structure. Experiments show that SuperCNN can robustly detect salient objects and outperforms the state-of-the-art methods on three benchmark datasets.

Subfields: Cognitive Neuroscience, Computer Vision and Pattern Recognition

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

Recent advances in convolutional neural networks (2015, background): “[170] derives a spatial kernel and a range kernel to produce two meaningful sequences as 1-D CNN inputs, to describe color uniqueness and color distribution respectively.”
Salient Object Detection in the Deep Learning Era: An In-Depth Survey (2019, methodology): “Commonly adopted processing units include super-pixels/patches [28], [56], [60], and generic object proposals [26], [27], [36], [68].”
Pooling Methods in Deep Neural Networks, a Review (2020, background): “…efficiency, super-pixel has become a midlevel image representation and is widely used in computer vision algorithms such as object detection [29, 30], semantic segmentation [31-34], saliency estimation [35-38], optical flow estimation [39, 40], depth estimation [41, 42], and object tracking [43].”
Stacked Cross Refinement Network for Edge-Aware Salient Object Detection (2019, background): “Early deep salient object detection models [14, 18, 19, 27, 38] utilize classi-ﬁcation network to determine each region of an image is salient or not.”
Deeply Supervised Salient Object Detection with Short Connections (2016, methodology): “[46] presented a superpixel-wise convolutional neural network architectures by utilizing hierarchical contrast features.”
Comprehensive Review of Artificial Neural Network Applications to Pattern Recognition (2019, background): “CNN is a group of deep feedforward artificial neural networks (DFFANNs), normally applied to analyze visual imagery [13], [14].”
Review of Visual Saliency Detection With Comprehensive Information (2018, methodology): “Some hierarchical deep networks for saliency detection are proposed, such as SuperCNN [38], and DHSNet [41].”
Superpixels: An evaluation of the state-of-the-art (2016, background): “We found that UE tends to correlate strongly with ASA which can be explained by Equations (3) and (6), respectively.”
Object Detection With Deep Learning: A Review (2018, methodology): “To learn internal representations of saliency efﬁciently, He et al. [140] proposed a novel superpixelwise CNN approach called SuperCNN, in which salient object detection is formulated as a binary labeling problem.”
RGBD Salient Object Detection via Deep Fusion (2016, methodology): “2) Performance Comparison: In this section, we compare our method with four state-of-the-art methods designed for RGB images (S-CNN [41], BSCA [16], MB+ [53], and LEGS [42]) and four RGBD saliency methods designed especially for RGBD images (LMH [12], ACSD [13], GP [14], and LBE [15]).”
New Contour Cue-Based Hybrid Sparse Learning for Salient Object Detection (2019, unknown): “Besides the above traditional approaches, deep learning has recently been adopted to extract salient features or directly learn saliency scores from raw images, such as LEGS [42], super-CNN [43], and SDAE [44].”
Salient Object Detection via Structured Matrix Decomposition (2017, background): “Most recently, multiple kernel learning [58] and convolutional neural network [59] techniques have been introduced to learn more robust discrimination between salient and non-salient regions.”