Filters








529 Hits in 4.1 sec

Deeply Supervised Rotation Equivariant Network for Lesion Segmentation in Dermoscopy Images [article]

Xiaomeng Li, Lequan Yu, Chi-Wing Fu, Pheng-Ann Heng
2018 arXiv   pre-print
The dermoscopy images exhibits rotational and reflectional symmetry, however, this geometric property has not been encoded in the state-of-the-art convolutional neural networks based skin lesion segmentation  ...  The whole framework is equivariant to input transformations, including rotation and reflection, which improves the network efficiency and thus contributes to the segmentation performance.  ...  We consider to improve the network efficiency by encoding the rotation and flipping equivariance into the network, in which the network preserves the equivariance inherent without relying on data augmentation  ... 
arXiv:1807.02804v1 fatcat:3ro2ahxxe5dedmbgxw5cnnb7cm

Beyond CNNs: Exploiting Further Inherent Symmetries in Medical Images for Segmentation [article]

Shuchao Pang, Anan Du, Mehmet A. Orgun, Yan Wang, Quanzheng Sheng, Shoujin Wang, Xiaoshui Huang, Zhemei Yu
2020 arXiv   pre-print
Although the existing methods based on convolutional neural networks (CNNs) have achieved the state-of-the-art performance, many challenges still remain in medical tumor segmentation.  ...  Then, to keep segmentation networks globally equivariant, we design distinctive group layers with layerwise symmetry constraints.  ...  Data Augmentation & Rotation Equivariant Networks Data augmentation is a widely used technique to train a more robust neural network model for real applications [Ciresan et al., 2013] .  ... 
arXiv:2005.03924v1 fatcat:pfpx24awcbfpfkbkxfr66zkuta

Beyond CNNs: Exploiting Further Inherent Symmetries in Medical Image Segmentation [article]

Shuchao Pang, Anan Du, Mehmet A. Orgun, Yan Wang, Quan Z. Sheng, Shoujin Wang, Xiaoshui Huang, Zhenmei Yu
2022 arXiv   pre-print
Although the existing methods based on Convolutional Neural Networks (CNNs) have achieved the state-of-the-art performance, many challenges still remain in medical tumor segmentation.  ...  Then, to keep segmentation networks globally equivariant, we design distinctive group layers with layer-wise symmetry constraints.  ...  Data Augmentation & Rotation Equivariant Networks Data augmentation is a widely used technique to train a more robust neural network model for real applications [6] .  ... 
arXiv:2207.14472v1 fatcat:advrih7varafhohc3glne3hxvm

Rotation Equivariant CNNs for Digital Pathology [article]

Bastiaan S. Veeling, Jasper Linmans, Jim Winkens, Taco Cohen, Max Welling
2018 arXiv   pre-print
We propose a new model for digital pathology segmentation, based on the observation that histopathology images are inherently symmetric under rotation and reflection.  ...  We present a visual analysis showing improved stability on predictions, and demonstrate that exploiting rotation equivariance significantly improves tumor detection performance on a challenging lymph node  ...  A proven approach is to use convolutional neural networks (CNNs), a type of deep learning model, trained on patches extracted from whole-slide images.  ... 
arXiv:1806.03962v1 fatcat:nijlmsxy2jbhpf3cir7hvdxa7q

Extracting Invariant Features From Images Using An Equivariant Autoencoder

Denis Kuzminykh, Daniil Polykovskiy, Alexander Zhebrak
2018 Asian Conference on Machine Learning  
Convolutional Neural Networks achieve state of the art results in many image recognition tasks.  ...  We apply group convolutions to build an Equivariant Autoencoder with embeddings that change predictably under the specified set of transformations.  ...  Cohen and Welling (2016) described group equivariant convolutional neural networks (G-CNNs) with G-convolution operator.  ... 
dblp:conf/acml/KuzminykhPZ18 fatcat:42wha2roqvbotoxuplke42hdi4

CubeNet: Equivariance to 3D Rotation and Translation [chapter]

Daniel Worrall, Gabriel Brostow
2018 Lecture Notes in Computer Science  
We introduce a Group Convolutional Neural Network with linear equivariance to translations and right angle rotations in three dimensions.  ...  3D Convolutional Neural Networks are sensitive to transformations applied to their input.  ...  Our V rand metric is slightly improved over UNet and Quan et al. , but not as good as Weiler et al. , who use a 2D group convolutional neural network approach, with 17 rotations about the z-axis and lifting  ... 
doi:10.1007/978-3-030-01228-1_35 fatcat:tslryczdm5hilkldstogjajlfa

Rotation Equivariant Deforestation Segmentation and Driver Classification [article]

Joshua Mitton, Roderick Murray-Smith
2021 arXiv   pre-print
In this work, we develop a rotation equivariant convolutional neural network model to predict the drivers and generate segmentation maps of deforestation events from Landsat 8 satellite images.  ...  This outperforms previous methods in classifying the drivers and predicting the segmentation map of deforestation, offering a 9% improvement in classification accuracy and a 7% improvement in segmentation  ...  A range of prior works have used decision trees, random forest classifiers, and convolutional neural networks for the task of classifying and mapping deforestation drivers (Phiri et al., 2019; Descals  ... 
arXiv:2110.13097v2 fatcat:brkfvdcrvbgvbdl76kwdyfvsxq

Learning Steerable Filters for Rotation Equivariant CNNs [article]

Maurice Weiler, Fred A. Hamprecht, Martin Storath
2018 arXiv   pre-print
Convolutional neural networks (CNNs) implement translational equivariance by construction; for other transformations, however, they are compelled to learn the proper mapping.  ...  We utilize group convolutions which guarantee an equivariant mapping.  ...  Introduction Convolutional neural networks are extremely successful predictive models when the input data has spatial structure.  ... 
arXiv:1711.07289v3 fatcat:de4jmbhvebc2ba75lslzl7nhhq

Learning Steerable Filters for Rotation Equivariant CNNs

Maurice Weiler, Fred A. Hamprecht, Martin Storath
2018 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition  
Convolutional neural networks (CNNs) implement translational equivariance by construction; for other transformations, however, they are compelled to learn the proper mapping.  ...  We utilize group convolutions which guarantee an equivariant mapping.  ...  Introduction Convolutional neural networks are extremely successful predictive models when the input data has spatial structure.  ... 
doi:10.1109/cvpr.2018.00095 dblp:conf/cvpr/WeilerHS18 fatcat:uwj34l7zpvfyjbunapwwdxx424

CubeNet: Equivariance to 3D Rotation and Translation [article]

Daniel Worrall, Gabriel Brostow
2018 arXiv   pre-print
We introduce a Group Convolutional Neural Network with linear equivariance to translations and right angle rotations in three dimensions.  ...  3D Convolutional Neural Networks are sensitive to transformations applied to their input.  ...  Our V rand metric is slightly improved over UNet and Quan et al. , but not as good as Weiler et al. , who use a 2D group convolutional neural network approach, with 17 rotations about the z-axis and lifting  ... 
arXiv:1804.04458v1 fatcat:zil6lp6b5bahpf6c6tkbjsojmy

A Data and Compute Efficient Design for Limited-Resources Deep Learning [article]

Mirgahney Mohamed, Gabriele Cesa, Taco S. Cohen, Max Welling
2020 arXiv   pre-print
Thanks to their improved data efficiency, equivariant neural networks have gained increased interest in the deep learning community.  ...  However, equivariant models are commonly implemented using large and computationally expensive architectures, not suitable to run on mobile devices.  ...  (2018) , we augment the training set with π 2 rotations and reflections.  ... 
arXiv:2004.09691v2 fatcat:vxshbh2g25hupjbl4qgp6nvdkm

Neural Ordinary Differential Equations for Semantic Segmentation of Individual Colon Glands [article]

Hans Pinckaers, Geert Litjens
2019 arXiv   pre-print
Automated medical image segmentation plays a key role in quantitative research and diagnostics. Convolutional neural networks based on the U-Net architecture are the state-of-the-art.  ...  Recently, Neural Ordinary Differential Equations (NODE) have been proposed, a new type of continuous depth deep neural network.  ...  The current de facto standard network architecture for segmentation in medical images are convolutional neural networks, specifically those following the U-Net architecture [6] .  ... 
arXiv:1910.10470v1 fatcat:5u3kgwv5fnax5myntslbeonu64

Front Matter: Volume 11313

Bennett A. Landman, Ivana Išgum
2020 Medical Imaging 2020: Image Processing  
The papers reflect the work and thoughts of the authors and are published herein as submitted.  ...  convolutional neural networks improve segmentation over reflection augmentation Proc. of SPIE Vol. 11313 1131301-9 Variational intensity cross channel encoder for unsupervised vessel segmentation on OCT  ...  networks for multi-organ segmentation with improved data augmentation and instance normalization 11313 17Identification of kernels in a convolutional neural network: connections between the level set  ... 
doi:10.1117/12.2570657 fatcat:be32besqknaybh6wibz7unuboa

Learning Invariances in Neural Networks from Training Data

Gregory W. Benton, Marc Finzi, Pavel Izmailov, Andrew Gordon Wilson
2020 Neural Information Processing Systems  
Invariances to translations have imbued convolutional neural networks with powerful generalization properties.  ...  We show how to learn invariances and equivariances by parameterizing a distribution over augmentations and optimizing the training loss simultaneously with respect to the network parameters and augmentation  ...  it is possible to learn that we want to use a convolutional neural network.  ... 
dblp:conf/nips/BentonFIW20 fatcat:cwn2sli4bfgu7iwxjemfoy2ony

Enhanced Rotation-Equivariant U-Net for Nuclear Segmentation

Benjamin Chidester, That-Vinh Ton, Minh-Triet Tran, Jian Ma, Minh N. Do
2019 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)  
Specifically, we consider enforcing rotation equivariance in the network, the placement of residual blocks, and applying novel data augmentation designed specifically for histopathology images, and show  ...  segmentation.  ...  A group-equivariant neural network can be created by the composition of group-equivariant convolution with several other group operations, given below, which preserve equivariance to such transformations  ... 
doi:10.1109/cvprw.2019.00143 dblp:conf/cvpr/ChidesterTT0D19 fatcat:lac56wjkuvhdbi4qcjzysqblv4
« Previous Showing results 1 — 15 out of 529 results