18 Hits in 4.5 sec

SRM : A Style-based Recalibration Module for Convolutional Neural Networks [article]

HyunJae Lee, Hyo-Eun Kim, Hyeonseob Nam
2019 arXiv   pre-print
We propose a Style-based Recalibration Module (SRM), a simple yet effective architectural unit, which adaptively recalibrates intermediate feature maps by exploiting their styles.  ...  Following the advance of style transfer with Convolutional Neural Networks (CNNs), the role of styles in CNNs has drawn growing attention from a broader perspective.  ...  Conclusion In this work, we present Style-based Recalibration Module (SRM), a lightweight architectural unit that dynamically recalibrates feature responses based on style importance.  ... 
arXiv:1903.10829v1 fatcat:nlnpezhzsraudctqd4d7qiybzq

Linking Multi-Layer Dynamical GCN With Style-Based Recalibration CNN for EEG-Based Emotion Recognition

Guangcheng Bao, Kai Yang, Li Tong, Jun Shu, Rongkai Zhang, Linyuan Wang, Bin Yan, Ying Zeng
2022 Frontiers in Neurorobotics  
Particularly, we add style-based recalibration module to CNN to extract deep layer features, which can better select features that are highly related to emotion.  ...  In this paper, we proposed a novel model combining graph convolutional network and convolutional neural network, namely MDGCN-SRCNN, aiming to fully extract features of channel connectivity in different  ...  LT and BY was mainly responsible for research design. JS was mainly responsible for data collection and production of charts. RZ was mainly responsible for production of charts.  ... 
doi:10.3389/fnbot.2022.834952 pmid:35280845 pmcid:PMC8907537 fatcat:j5lhgcivwfgjbklvvqa7rkysdu

SAST-GCN: Segmentation Adaptive Spatial Temporal-Graph Convolutional Network for P3-Based Video Target Detection

Runnan Lu, Ying Zeng, Rongkai Zhang, Bin Yan, Li Tong
2022 Frontiers in Neuroscience  
Especially, a style-based recalibration module is added to select feature maps with higher contributions and increase the feature extraction ability of the network.  ...  This paper proposes a segmentation adaptive spatial-temporal graph convolutional network (SAST-GCN) for P3-based video target detection.  ...  were mainly responsible for research design. LT was mainly responsible for data collection and manuscript modification. All authors contributed to the article and approved the submitted version.  ... 
doi:10.3389/fnins.2022.913027 pmid:35720707 pmcid:PMC9201684 fatcat:apgatuvt7jdrdkwrrhwzczaghy

Deep Reinforced Attention Learning for Quality-Aware Visual Recognition [article]

Duo Li, Qifeng Chen
2022 arXiv   pre-print
Given an existing neural network equipped with arbitrary attention modules, we introduce a meta critic network to evaluate the quality of attention maps in the main network.  ...  In this paper, we build upon the weakly-supervised generation mechanism of intermediate attention maps in any convolutional neural networks and disclose the effectiveness of attention modules more straightforwardly  ...  Style Recalibration. Recently it is revealed that the style information also plays an important role in the decision process of neural networks.  ... 
arXiv:2007.06156v2 fatcat:lx5juk5in5eufagtadav657j2e

Calibrated Convolution with Gaussian of Difference

Huoxiang Yang, Chao Li, Yongsheng Liang, Wei Liu, Fanyang Meng
2022 Applied Sciences  
Attention mechanisms are widely used for Convolutional Neural Networks (CNNs) when performing various visual tasks.  ...  A simple yet effective scale-invariant attention module that operates within a single convolution is able to adaptively build powerful scale-invariant features to recalibrate the feature representation  ...  Introduction Convolutional Neural Networks (CNNs) have proven to be effective at tackling a wide range of visual tasks [1, 2] .  ... 
doi:10.3390/app12136570 fatcat:cd4owkn325dw5hjw7v35otdzuq

SimAM: A Simple, Parameter-Free Attention Module for Convolutional Neural Networks

Lingxiao Yang, Ru-Yuan Zhang, Lida Li, Xiaohua Xie
2021 International Conference on Machine Learning  
In this paper, we propose a conceptually simple but very effective attention module for Convolutional Neural Networks (ConvNets).  ...  In contrast to existing channel-wise and spatial-wise attention modules, our module instead infers 3-D attention weights for the feature map in a layer without adding parameters to the original networks  ...  Acknowledgements We thank all reviewers for their kindly and constructive suggestions.  ... 
dblp:conf/icml/YangZLX21 fatcat:4wqmdd7fczenxdzhygrycpjsje

An Efficient Lightweight SAR Ship Target Detection Network with Improved Regression Loss Function and Enhanced Feature Information Expression

Jimin Yu, Tao Wu, Xin Zhang, Wei Zhang
2022 Sensors  
Thereby, an efficient lightweight network Efficient-YOLO for ship detection in complex situations is proposed in the present work.  ...  It is difficult to identify the ship images obtained by a synthetic aperture radar (SAR) due to the influence of dense ships, complex background and small target size, so a deep learning-based target detection  ...  Related methods Style-Based Recalibration (SRM) Module The SRM [22] is a simple and efficient architecture unit that adaptively recalibrates feature information through the style of intermediate feature  ... 
doi:10.3390/s22093447 pmid:35591135 pmcid:PMC9101316 fatcat:uze2oavbfffzbpw5ri4yioeazy

Attention Mechanisms in Computer Vision: A Survey [article]

Meng-Hao Guo, Tian-Xing Xu, Jiang-Jiang Liu, Zheng-Ning Liu, Peng-Tao Jiang, Tai-Jiang Mu, Song-Hai Zhang, Ralph R. Martin, Ming-Ming Cheng, Shi-Min Hu
2021 arXiv   pre-print
Such an attention mechanism can be regarded as a dynamic weight adjustment process based on features of the input image.  ...  We also suggest future directions for attention mechanism research.  ...  We would like to thank Cheng-Ze Lu, Zhengyang Geng, Shilong liu, He Wang, Huiying Lu and Chenxi Huang for their helpful discussions and insightful suggestions.  ... 
arXiv:2111.07624v1 fatcat:2nreclzam5eqhmwolot5fkhvyu

EPSANet: An Efficient Pyramid Squeeze Attention Block on Convolutional Neural Network [article]

Hu Zhang and Keke Zu and Jian Lu and Yuru Zou and Deyu Meng
2021 arXiv   pre-print
Recently, it has been demonstrated that the performance of a deep convolutional neural network can be effectively improved by embedding an attention module into it.  ...  By replacing the 3x3 convolution with the PSA module in the bottleneck blocks of the ResNet, a novel representational block named Efficient Pyramid Squeeze Attention (EPSA) is obtained.  ...  By embedding a operator for multi-scale feature extraction into a convolution neural network(CNN), a more effective feature representation ability can be obtained.  ... 
arXiv:2105.14447v2 fatcat:syxxgnug3rggbhhpxjhvmowve4

Sequential Feature Filtering Classifier [article]

Minseok Seo, Jaemin Lee, Jongchan Park, Dong-Geol Choi
2020 arXiv   pre-print
We propose Sequential Feature Filtering Classifier (FFC), a simple but effective classifier for convolutional neural networks (CNNs).  ...  The sequential feature filtering process generates multiple features, which are fed into a shared classifier for multiple outputs.  ...  Lastly, the recently proposed Style-based Recalibration Module (SRM) improved the performance of CNN by proposing a channel-independent style integration method utilizing style pooling.  ... 
arXiv:2006.11808v1 fatcat:hmn2ctq2ifapjn5his5fqggltq

Studying the Effects of Self-Attention for Medical Image Analysis [article]

Adrit Rao, Jongchan Park, Sanghyun Woo, Joon-Young Lee, Oliver Aalami
2021 arXiv   pre-print
However, the standard convolutional neural network (CNN) does not necessarily employ subconscious feature relevancy evaluation techniques similar to the trained medical specialist and evaluates features  ...  Through both quantitative and qualitative evaluations along with a clinical user-centric survey study, we aim to provide a deeper understanding of the effects of self-attention in medical computer vision  ...  The Style-based Re-calibration Module (SRM) [23] is a simple yet powerful channel attention module that accounts for channel statistics (mean and standard deviation) when scaling the channel values.  ... 
arXiv:2109.01486v1 fatcat:munlozczivgnxd2555brx2zyse

Sequential Feature Filtering Classifier

Minseok Seo, Jaemin Lee, Jongchan Park, Daehan Kim, Dong-Geol Choi
2021 IEEE Access  
Lastly, the recently proposed Style-based Recalibration Module (SRM) improved the performance of CNN by proposing a channelindependent style integration method utilizing style pooling.  ...  Squeeze-and-Excitation (SE) improved the performance of CNN through a channel-wise recalibration operator, and the Convolutional block attention module (CBAM) achieved a further improved performance by  ... 
doi:10.1109/access.2021.3090439 fatcat:oikeol6rgrho5medgxyeer35im

Reducing Domain Gap by Reducing Style Bias [article]

Hyeonseob Nam, HyunJae Lee, Jongchan Park, Wonjun Yoon, Donggeun Yoo
2021 arXiv   pre-print
Convolutional Neural Networks (CNNs) often fail to maintain their performance when they confront new test domains, which is known as the problem of domain shift.  ...  Our Style-Agnostic Networks (SagNets) disentangle style encodings from class categories to prevent style biased predictions and focus more on the contents.  ...  BIN [40] improved classification performance by reducing unnecessary style information using trainable normalization, and SRM [28] extended this idea to style-based feature recalibration.  ... 
arXiv:1910.11645v4 fatcat:ugujqj3jxrffjicyz5oouv5gna

Learning Visual Context by Comparison [article]

Minchul Kim, Jongchan Park, Seil Na, Chang Min Park, Donggeun Yoo
2020 arXiv   pre-print
In this paper, we present Attend-and-Compare Module (ACM) for capturing the difference between an object of interest and its corresponding context.  ...  For evaluation, we apply our module to three chest X-ray recognition tasks and COCO object detection & segmentation tasks and observe consistent improvements across tasks.  ...  A Style-based Recalibration Module (SRM) [22] further explores the global feature modeling in terms of style recalibration.  ... 
arXiv:2007.07506v1 fatcat:mhwvyzjvm5edzlrzkueg4yepiy

Table of Contents

2019 2019 IEEE/CVF International Conference on Computer Vision (ICCV)  
Université de Montréal) SRM: A Style-Based Recalibration Module for Convolutional Neural Networks 1854 Hyunjae Lee(Lunit Inc.),Hyo-Eun Kim (Lunit Inc.), and Hyeonseob Nam(Lunit Inc.)  ...  Features 3394 Juhong Min (POSTECH), Jongmin Lee (POSTECH), Jean Ponce (Inria), and Minsu Cho (POSTECH) Information Entropy Based Feature Pooling for Convolutional Neural Networks 3404 Weitao Wan (Tsinghua  ... 
doi:10.1109/iccv.2019.00004 fatcat:5aouo4scprc75c7zetsimylj2y
« Previous Showing results 1 — 15 out of 18 results