Filters








341 Hits in 3.8 sec

Saliency Detection by Forward and Backward Cues in Deep-CNNs [article]

Nevrez Imamoglu, Chi Zhang, Wataru Shimoda, Yuming Fang, Boxin Shi
2017 arXiv   pre-print
The model detects attentive regions based on their objectness scores predicted by selected features from CNNs.  ...  As the proposed model is an effective integration of forward and backward cues through objectness without any supervision or regression to ground truth data, it gives promising results compared to state-of-the-art  ...  Eq.3 denotes the integration of forward (FWLayer_n) and backward (BWLayer_n) saliency cues from sub-model layers.  ... 
arXiv:1703.00152v2 fatcat:6kqspoaz4bdtzoefcx3zlbwz2y

Excitation Backprop for RNNs

Sarah Adel Bargal, Andrea Zunino, Donghyun Kim, Jianming Zhang, Vittorio Murino, Stan Sclaroff
2018 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition  
In this work, we devise a formulation that simultaneously grounds evidence in space and time, in a single pass, using top-down saliency.  ...  Our model employs a single backward pass to produce saliency maps that highlight the evidence that a given RNN used in generating its outputs.  ...  This work was supported in part by NSF grants 1551572 and 1029430, an IBM PhD Fellowship, gifts from Adobe and NVidia, and Intelligence Advanced Research Projects Activity (IARPA) via Department of Interior  ... 
doi:10.1109/cvpr.2018.00156 dblp:conf/cvpr/BargalZKZMS18 fatcat:oqv3nyo52fbehd3ic3cuvsn5wy

Excitation Backprop for RNNs [article]

Sarah Adel Bargal, Andrea Zunino, Donghyun Kim, Jianming Zhang, Vittorio Murino, Stan Sclaroff
2018 arXiv   pre-print
In this work, we devise a formulation that simultaneously grounds evidence in space and time, in a single pass, using top-down saliency.  ...  Grounding decisions made by deep networks has been studied in spatial visual content, giving more insight into model predictions for images.  ...  This work was supported in part by NSF grants 1551572 and 1029430, an IBM PhD Fellowship, gifts from Adobe and NVidia, and Intelligence Advanced Research Projects Activity (IARPA) via Department of Interior  ... 
arXiv:1711.06778v3 fatcat:io6onint6zbcvc6b5asfazrnee

Video Saliency Prediction Using Enhanced Spatiotemporal Alignment Network [article]

Jin Chen, Huihui Song, Kaihua Zhang, Bo Liu, Qingshan Liu
2020 arXiv   pre-print
The output of MDAN is then fed into the Bi-ConvLSTM for further enhancement, which captures the useful long-time temporal information along forward and backward timing directions to effectively guide attention  ...  The MDAN learns to align the features of the neighboring frames to the reference one in a coarse-to-fine manner, which can well handle various motions.  ...  Mr-CNN [31] employs a multi-resolution CNN guided by both bottom-up visual saliency and top-down visual cues to predict visual saliency. Figure 1 1 shows an overview of the proposed model for VSP.  ... 
arXiv:2001.00292v1 fatcat:mu5m34vx2bbyzfgmvs6uqeux2e

Video Saliency Detection Using Bi-directional LSTM

2020 KSII Transactions on Internet and Information Systems  
We combine the Convolutional Neural Network (CNN) and the Deep Bidirectional LSTM Network (DB-LSTM) to learn the spatio-temporal features by exploring the object motion information and object motion information  ...  Deep learning can extract the edge features of the image, providing technical support for video saliency. This paper proposes a new detection method.  ...  Developed CNN and DB-LSTM network structures, and innovatively developed bidirectional LSTM, enabling CNN extracted features to be cascaded forward and backward in LSTM, whether for intra-frame significance  ... 
doi:10.3837/tiis.2020.06.007 fatcat:qganemcvzvh7hnu4go2svrxxqm

An integration of bottom-up and top-down salient cues on RGB-D data: saliency from objectness versus non-objectness

Nevrez Imamoglu, Wataru Shimoda, Chi Zhang, Yuming Fang, Asako Kanezaki, Keiji Yanai, Yoshifumi Nishida
2017 Signal, Image and Video Processing  
In this work, we combine bottom-up and top-down cues from both space and object based salient features on RGB-D data.  ...  Saliency models generally incorporate salient cues either in bottom-up or top-down norm separately.  ...  Our experiments demonstrate that these feed-forward computations from fully supervised models [27, 28] or backward process (gradients from back-propagation for CNN layers) from weakly supervised models  ... 
doi:10.1007/s11760-017-1159-7 fatcat:b5nr2bj7bzbbzb7nmkdrmnseiq

Deeply-Supervised Recurrent Convolutional Neural Network for Saliency Detection

Youbao Tang, Xiangqian Wu, Wei Bu
2016 Proceedings of the 2016 ACM on Multimedia Conference - MM '16  
This paper proposes a novel saliency detection method by developing a deeply-supervised recurrent convolutional neural network (DSRCNN), which performs a full image-to-image saliency prediction.  ...  For saliency detection, the local, global, and contextual information of salient objects is important to obtain a high quality salient map.  ...  A large number of visual saliency detection approaches have been proposed by exploiting different salient cues recently.  ... 
doi:10.1145/2964284.2967250 dblp:conf/mm/TangWB16 fatcat:xjobladncze45luax2evr6m3sq

Deeply-Supervised Recurrent Convolutional Neural Network for Saliency Detection [article]

Youbao Tang, Xiangqian Wu, Wei Bu
2016 arXiv   pre-print
This paper proposes a novel saliency detection method by developing a deeply-supervised recurrent convolutional neural network (DSRCNN), which performs a full image-to-image saliency prediction.  ...  For saliency detection, the local, global, and contextual information of salient objects is important to obtain a high quality salient map.  ...  A large number of visual saliency detection approaches have been proposed by exploiting different salient cues recently.  ... 
arXiv:1608.05177v1 fatcat:4yzximiusrhwtiq3zcincmpgn4

Spatial Audio Feature Discovery with Convolutional Neural Networks

Etienne Thuillier, Hannes Gamper, Ivan J. Tashev
2018 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)  
These maps, in addition to the convolution filters learned by the CNN, are discussed in the context of listening tests reported in the literature.  ...  LRP provides saliency maps that can be used to identify spectral features used by the network for classification.  ...  ., its relevance R, from network output and backwards, layer-by-layer, to network input.  ... 
doi:10.1109/icassp.2018.8462315 dblp:conf/icassp/ThuillierGT18 fatcat:dhzyvf4qtna6riaygbzkk6xqre

Attention Embedded Spatio-Temporal Network for Video Salient Object Detection

Lili Huang, Pengxiang Yan, Guanbin Li, Qing Wang, Liang Lin
2019 IEEE Access  
The main challenge in video salient object detection is how to model object motion and dramatic changes in appearance contrast.  ...  Therefore, using the flow-guided attention map alone causes the spatial saliency to be influenced by all moving objects rather than just the salient objects, resulting in unstable and temporally inconsistent  ...  However, both the forward and backward frames in a video are considerable and complementary in video saliency prediction.  ... 
doi:10.1109/access.2019.2953046 fatcat:rciq4rmf5nhc5icsvkdmlqdchq

Video Salient Object Detection via Fully Convolutional Networks

Wenguan Wang, Jianbing Shen, Ling Shao
2018 IEEE Transactions on Image Processing  
This paper proposes a deep learning model to efficiently detect salient regions in videos.  ...  It addresses two important issues: (1) deep video saliency model training with the absence of sufficiently large and pixel-wise annotated video data, and (2) fast video saliency training and detection.  ...  Inspired by this, we investigate CNNs to another computer vision task, namely video saliency detection.  ... 
doi:10.1109/tip.2017.2754941 pmid:28945593 fatcat:v644yvm4qjag5l5ri5lq7ztwee

Bidirectional LSTM with saliency-aware 3D-CNN features for human action recognition

Sheeraz Arif, Department of Information and Communication Engineering, School of Information and Electronics, Beijing Institute of Technology, Beijing, 100081, China, Jing Wang, Adnan Ahmed Siddiqui, Rashid Hussain, Fida Hussain, Department of Information and Communication Engineering, School of Information and Electronics, Beijing Institute of Technology, Beijing, 100081, China, Department of Computing, Faculty of Engineering Science and Technology, Hamdard University, Karachi, Pakistan, Department of Computing, Faculty of Engineering Science and Technology, Hamdard University, Karachi, Pakistan, School of Electrical and Information Engineering, Jiangsu University, Nanjing, China
2021 Maǧallaẗ al-abḥāṯ al-handasiyyaẗ  
First, we generate a saliency-aware video stream by applying the saliency-aware method.  ...  ., RGB stream and saliency-aware video stream, to collect both spatial and semantic temporal features. Next, a deep bidirectional LSTM network is used to learn sequential deep temporal dynamics.  ...  The combined output is calculated based on hidden states of both backward and forward directions.  ... 
doi:10.36909/jer.v9i3a.8383 fatcat:55whmd65lfh2zp4tob5gjpspay

Unsupervised motion saliency map estimation based on optical flow inpainting [article]

L. Maczyta, P. Bouthemy, O. Le Meur
2019 arXiv   pre-print
The residual flow in these regions is given by the difference between the optical flow and the flow inpainted from the surrounding areas. It provides the cue for motion saliency.  ...  The method is flexible and general by relying on motion information only.  ...  Fig. 1 . 1 Overall framework of our method for motion saliency map estimation with the two backward and forward streams.  ... 
arXiv:1903.04842v1 fatcat:oext6xyjoncdxnahnmie5jlrbu

Detect Globally, Refine Locally: A Novel Approach to Saliency Detection

Tiantian Wang, Lihe Zhang, Shuo Wang, Huchuan Lu, Gang Yang, Xiang Ruan, Ali Borji
2018 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition  
They simply apply concatenation or element-wise operation to incorporate high-level semantic cues and low-level detailed information.  ...  To address this problem, we proposes a global Recurrent Localization Network (RLN) which exploits contextual information by the weighted response map in order to localize salient objects more accurately  ...  Figure 4 illustrates the overall recurrent structure in the process of forward-and backward-propagation following depth and time dimensions (here we set t = 1).  ... 
doi:10.1109/cvpr.2018.00330 dblp:conf/cvpr/WangZWL0RB18 fatcat:w2ulbiyv75e6ti2iqt5r7ipdpi

Deep Visual Attention Prediction

Wenguan Wang, Jianbing Shen
2018 IEEE Transactions on Image Processing  
Although Convolutional Neural Networks (CNNs) have made substantial improvement on human attention prediction, it is still needed to improve CNN based attention models by efficiently leveraging multi-scale  ...  Our model is learned in a deep supervision manner, where supervision is directly fed into multi-level layers, instead of previous approaches of providing supervision only at the output layer and propagating  ...  It has been proved that saliency cues on different level and scales are important in saliency detection [63] , [37] .  ... 
doi:10.1109/tip.2017.2787612 pmid:29990140 fatcat:2h2r55lu7vacjfawqdcm6e7bee
« Previous Showing results 1 — 15 out of 341 results