41 Hits in 4.2 sec

Attention W-Net: Improved Skip Connections for better Representations [article]

Shikhar Mohan, Saumik Bhattacharya, Sayantari Ghosh
2021 arXiv   pre-print
We propose Attention W-Net, a new U-Net based architecture for retinal vessel segmentation to address these problems.  ...  Our Attention Block uses decoder features to attend over the encoder features from skip-connections during upsampling, resulting in higher compatibility when the encoder and decoder features are added.  ...  with skip connections which allow for an efficient flow of gradients for learning.  ... 
arXiv:2110.08811v1 fatcat:5m55lijy5vekjkz52ermumwmxu

W-Net: A Deep Model for Fully Unsupervised Image Segmentation [article]

Xide Xia, Brian Kulis
2017 arXiv   pre-print
While significant attention has been recently focused on designing supervised deep semantic segmentation algorithms for vision tasks, there are many domains in which sufficient supervised pixel-level labels  ...  We borrow recent ideas from supervised semantic segmentation methods, in particular by concatenating two fully convolutional networks together into an autoencoder--one for encoding and one for decoding  ...  Further, Figure [11] and Figure [12] illustrates more results of running the W-Net+ucm on images from the BSDS500.  ... 
arXiv:1711.08506v1 fatcat:ev74kdmwynhgpdkk7z4u2vr62a

A Novel Squeeze-and-Excitation W-Net for 2D and 3D Building Change Detection with Multi-Source and Multi-Feature Remote Sensing Data

Haiming Zhang, Mingchang Wang, Fengyan Wang, Guodong Yang, Ying Zhang, Junqian Jia, Siqi Wang
2021 Remote Sensing  
In this article, we propose a novel and general squeeze-and-excitation W-Net, which is developed from U-Net and SE-Net.  ...  The experimental results in two 2D data sets and two challenging 3D data sets demonstrate that the promising performances of the squeeze-and-excitation W-Net outperform several traditional and state-of-the-art  ...  We also would like to thank the School of Geomatics and Prospecting Engineering, Jilin Jianzhu University, for providing us with the LiDAR point cloud data.  ... 
doi:10.3390/rs13030440 fatcat:yvzswlo3yncrrfojoivijbl6pi

Adversarial Networks for Scale Feature-Attention Spectral Image Reconstruction from a Single RGB

Pengfei Liu, Huaici Zhao
2020 Sensors  
To provide a more accurate solution, we propose another distinct architecture, named W-Net, that builds one more branch compared to U-Net to conduct boundary supervision.  ...  We establish the feature pyramid inside the network and use the attention mechanism for feature selection.  ...  The feature pyramid and attention mechanism inside the network for feature selection improves the accuracy of hyperspectral reconstruction. (3) We further designed the W-Net structure based on SAPUNet  ... 
doi:10.3390/s20082426 pmid:32344686 pmcid:PMC7219499 fatcat:h4gcdz6zsjbo3bzbbj7f75ifma

AGCDetNet: An Attention-guided Network for Building Change Detection in High-resolution Remote Sensing Images

Kaiqiang Song, Jie Jiang
2021 IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing  
AGCDetNet learns to enhance the feature representation of change information and achieve accuracy improvements using spatial-and channelattention.  ...  Channel-wise attention-guided interference filtering unit (CIFU) / atrous spatial pyramid pooling module (CG-ASPP) enhances the representation of multilevel features and multiscale context, respectively  ...  (h) W-Net. (i) FC-EF-Res. (j) Peng et al. Best viewed in color. Fig. 8 . 8 CD results on the WHU dataset. Zoom in for an improved view.  ... 
doi:10.1109/jstars.2021.3077545 fatcat:emqmdfxegrcd3hxpafxraw4jtu

Feature pyramid U‐Net for retinal vessel segmentation

Yi‐Peng Liu, Xue Rui, Zhanqing Li, Dongxu Zeng, Jing Li, Peng Chen, Ronghua Liang
2021 IET Image Processing  
In this representation, objects features with different size like micro-vessels and pathology will be fused for better vessel segmentation.  ...  Therefore, it is of great significance to study effective retinal vessel segmentation methods and assist doctors in early diagnoses with quantitative results for vascular networks.  ...  FUNDING INFORMATION Natural Science Foundation of China (62076220, 61502426); Zhejiang Provincial Natural Science Foundation (LQ19H030004); the Fundamental Research Funds for the Provincial Universities  ... 
doi:10.1049/ipr2.12142 fatcat:bef6etvkljbhrpf3qwmkbrtj4y

CAggNet: Crossing Aggregation Network for Medical Image Segmentation [article]

Xu Cao, Yanghao Lin
2020 arXiv   pre-print
In CAggNet, the simple skip connection structure of general U-Net is replaced by aggregations of multi-level down-sampling and up-sampling layers, which is a new form of nested skip connection.  ...  In this paper, we present Crossing Aggregation Network (CAggNet), a novel densely connected semantic segmentation approach for medical image analysis.  ...  ACKNOWLEDGEMENTS We gratefully acknowledge the support of School of Data Science and School of Basic Medical Sciences at Fudan University for providing NVIDIA 1080Ti GPUs used for this research.  ... 
arXiv:2004.08237v2 fatcat:n5j2ngiegzdkpn3rhkylzo45ya

Difference Enhancement and Spatial-Spectral Non-Local Network for Change Detection in VHR Remote Sensing Images

Tao Lei, Jie Wang, Hailong Ning, Xingwu Wang, Dinghua Xue, Qi Wang, Asoke K. Nandi
2021 IEEE Transactions on Geoscience and Remote Sensing  
The W-Net [73] changes the pooling to convolution with a stride of 2 to avoid too much information loss, and merges features from dual networks for skip-connection.  ...  Because the skip-connection performed on difference images can provide more detailed information to resolve the erroneous attention.  ... 
doi:10.1109/tgrs.2021.3134691 fatcat:bfsueu4vy5ctfie5vcna2xrxym

BiO-Net: Learning Recurrent Bi-directional Connections for Encoder-Decoder Architecture [article]

Tiange Xiang, Chaoyi Zhang, Dongnan Liu, Yang Song, Heng Huang, Weidong Cai
2020 arXiv   pre-print
Our proposed bi-directional skip connections can be directly adopted into any encoder-decoder architecture to further enhance its capabilities in various task domains.  ...  Previous extensions of U-Net have focused mainly on the modification of its existing building blocks or the development of new functional modules for performance gains.  ...  W-Net [25] modifies U-Net to tackle unsupervised segmentation problem by concatenating two U-Nets via an autoencoder style model.  ... 
arXiv:2007.00243v2 fatcat:vstikcsb2bgzphnrcshsunjcwy

Self-Supervised Learning from Unlabeled Fundus Photographs Improves Segmentation of the Retina [article]

Jan Kukačka, Anja Zenz, Marcel Kollovieh, Dominik Jüstel, Vasilis Ntziachristos
2021 arXiv   pre-print
Fundus photography is the primary method for retinal imaging and essential for diabetic retinopathy prevention.  ...  Automated segmentation of fundus photographs would improve the quality, capacity, and cost-effectiveness of eye care screening programs.  ...  We repeated the experiments several times (N=4 for image segmentation, N=12 for domain transfer) and paired the results from matching training/validation splits.  ... 
arXiv:2108.02798v1 fatcat:fcmedcintnechhbxybt44lhfhy

Robust Deep Graph Based Learning for Binary Classification [article]

Minxiang Ye, Vladimir Stankovic, Lina Stankovic, Gene Cheung
2019 arXiv   pre-print
To penalize samples around the decision boundary, we propose two regularized loss functions for semi-supervised learning.  ...  Convolutional neural network (CNN)-based feature learning has become state of the art, since given sufficient training data, CNN can significantly outperform traditional methods for various classification  ...  W-Net For assigning edge weights W r to the graph G r , we first employ a CNN, denoted by CNN C r , to learn a deep metric function.  ... 
arXiv:1912.03321v1 fatcat:sxoqsqn35bfzdehitoddaoetri

Optimisation of 2D U-Net Model Components for Automatic Prostate Segmentation on MRI

Indriani P. Astono, James S. Welsh, Stephan Chalup, Peter Greer
2020 Applied Sciences  
For combining feature maps in each convolution block, it is only beneficial if a skip connection with concatenation is used.  ...  We found that for upsampling, the combination of interpolation and convolution is better than the use of transposed convolution.  ...  For phase 2, we considered the use of a skip connection in each convolution block to improve the performance of the models in phase 1 (i.e., UNet_S, UNet_S1 and UNet_S2).  ... 
doi:10.3390/app10072601 fatcat:imm5vlupjvg6ni52hgjwmupxlm

BU-Net: Brain Tumor Segmentation Using Modified U-Net Architecture

Mobeen Ur Rehman, SeungBin Cho, Jee Hong Kim, Kil To Chong
2020 Electronics  
The semantic segmentation of a brain tumor is of paramount importance for its treatment and prevention.  ...  The contextual information is extracted with the aggregating features to get better segmentation performance.  ...  W-Net is another architecture that resembles the U-Net architecture, which uses two-stage U-Net.  ... 
doi:10.3390/electronics9122203 fatcat:o7yunq6tf5c77mewwgjzesawgy

Towards Bi-directional Skip Connections in Encoder-Decoder Architectures and Beyond [article]

Tiange Xiang, Chaoyi Zhang, Xinyi Wang, Yang Song, Dongnan Liu, Heng Huang, Weidong Cai
2022 arXiv   pre-print
skip connections.  ...  The ineffective skip connections are then discarded to reduce computational costs and speed up network inference.  ...  W-Net (Xia and Kulis, 2017) concatenates two U-Nets head-to-tail to approach image segmentation tasks in an unsupervised style.  ... 
arXiv:2203.05709v2 fatcat:5e5223ui2zetfbitpkka777xrq

PL-Net: Progressive Learning Network for Medical Image Segmentation [article]

Junlong Cheng, Chengrui Gao, Chaoqing Wang, Zhangqiang Ming, Yong Yang, Min Zhu
2021 arXiv   pre-print
In recent years, segmentation methods based on deep convolutional neural networks (CNNs) have made state-of-the-art achievements for many medical analysis tasks.  ...  However, most of these approaches improve performance by optimizing the structure or adding new functional modules of the U-Net, which ignoring the complementation and fusion of the coarse-grained and  ...  W-Net [16] draws on the idea of the supervised semantic segmentation method, which solves the problem of unsupervised segmentation by connecting two U-Nets through an auto-encoder style model.  ... 
arXiv:2110.14484v1 fatcat:ojhg5tmgwrba5kq6mbzprf3lsy
« Previous Showing results 1 — 15 out of 41 results