Filters








85 Hits in 6.0 sec

Exploring the Landscape of Spatial Robustness [article]

Logan Engstrom, Brandon Tran, Dimitris Tsipras, Ludwig Schmidt, Aleksander Madry
2019 arXiv   pre-print
In this work, we thoroughly investigate the vulnerability of neural network--based classifiers to rotations and translations.  ...  However, state-of-the-art models turn out to be also vulnerable to other, more natural classes of perturbations such as translations and rotations.  ...  Sloan Research Fellowship, a Google Research Award, and the NSF grants CCF-1553428 and CNS-1815221.  ... 
arXiv:1712.02779v4 fatcat:2eing5ri2nhtddfy2pmlkcbnly

CNN Architectures for Geometric Transformation-Invariant Feature Representation in Computer Vision: A Review

Alhassan Mumuni, Fuseini Mumuni
2021 SN Computer Science  
While convolutional neural networks (CNNs) have inherent representation power that provides a high degree of invariance to geometric image transformations, they are unable to satisfactorily handle nontrivial  ...  Using these methods, it is possible to develop task-oriented solutions to deal with nontrivial transformations.  ...  Jiang and Mei [146] proposed a rotation-invariant CNN with a dedicated polar transformation layer that can be inserted into to learn rotation invariance.  ... 
doi:10.1007/s42979-021-00735-0 fatcat:3zrkaan7dncoja4e32u7jgwo4m

Advbox: a toolbox to generate adversarial examples that fool neural networks [article]

Dou Goodman, Hao Xin, Wang Yang, Wu Yuesheng, Xiong Junfeng, Zhang Huan
2020 arXiv   pre-print
Advbox is a toolbox to generate adversarial examples that fool neural networks in PaddlePaddle, PyTorch, Caffe2, MxNet, Keras, TensorFlow and it can benchmark the robustness of machine learning models.  ...  Small and often imperceptible perturbations to the input images are sufficient to fool the most powerful neural networks.  ...  A rotation and a translation suffice: Fooling cnns with simple transformations. arXiv preprint arXiv:1712.02779, 2017.  ... 
arXiv:2001.05574v5 fatcat:dqq3jde7lngdpnw5hrjfv35skm

Do CNNs Encode Data Augmentations? [article]

Eddie Yan, Yanping Huang
2021 arXiv   pre-print
Our approach uses features in pre-trained vision models with minimal additional processing to predict common properties transformed by augmentation (scale, aspect ratio, hue, saturation, contrast, and  ...  A fundamental question is whether neural network features encode data augmentation transformations.  ...  Ideally, vision models for these tasks would be equivariant to perturbations in color, translation, scale, and rotation.  ... 
arXiv:2003.08773v3 fatcat:ultgzwmrondtvimnh5gucqvlma

Truly shift-equivariant convolutional neural networks with adaptive polyphase upsampling [article]

Anadi Chaman, Ivan Dokmanić
2021 arXiv   pre-print
We address this problem by proposing adaptive polyphase upsampling (APS-U), a non-linear extension of conventional upsampling, which allows CNNs with symmetric encoder-decoder architecture (for example  ...  With MRI and CT reconstruction experiments, we show that networks containing APS-D/U layers exhibit state of the art equivariance performance without sacrificing on image reconstruction quality.  ...  Conference on Machine Learn- translation suffice: Fooling CNNs with simple transfor- ing. 09–15 Jun 2019, vol. 97 of Proceedings of Machine mations,” 2019.  ... 
arXiv:2105.04040v3 fatcat:64qp7x4dn5c7xfxalo3z52i3l4

The Effectiveness of Data Augmentation in Image Classification using Deep Learning [article]

Luis Perez, Jason Wang
2017 arXiv   pre-print
Previous work has demonstrated the effectiveness of data augmentation through simple techniques, such as cropping, rotating, and flipping input images.  ...  One of the more successful data augmentations strategies is the traditional transformations mentioned above. We also experiment with GANs to generate images of different styles.  ...  We hypothesis that a simple CNN already performs so well on MNIST so that neural augmentation provides no benefits.  ... 
arXiv:1712.04621v1 fatcat:2zczxxyf6bapfdn2s76u3zymzu

MimicGAN: Robust Projection onto Image Manifolds with Corruption Mimicking [article]

Rushil Anirudh, Jayaraman J. Thiagarajan, Bhavya Kailkhura, Timo Bremer
2020 arXiv   pre-print
Such corruptions are common in the real world, for example images in the wild come with unknown crops, rotations, missing pixels, or other kinds of non-linear distributional shifts which break current  ...  As a result, they have been widely adopted across a variety of applications, ranging from challenging inverse problems like image completion, to problems such as anomaly detection and adversarial defense  ...  training data, for e.g. shifts like simple corruptions, missing pixels, translations, rotation or scale changes etc.  ... 
arXiv:1912.07748v3 fatcat:danggesoinburbth2pixzj74pm

Fooling the primate brain with minimal, targeted image manipulation [article]

Li Yuan, Will Xiao, Giorgia Dellaferrera, Gabriel Kreiman, Francis E.H. Tay, Jiashi Feng, Margaret S. Livingstone
2022 arXiv   pre-print
Our results represent a valuable step in quantifying and characterizing the differences in perturbation robustness of biological and artificial vision.  ...  We generated 'deceptive images' of human faces, monkey faces, and noise patterns so that they are perceived as a different, pre-specified target category, and measured both monkey neuronal responses and  ...  Schade for discussion and assistance during the work.  ... 
arXiv:2011.05623v3 fatcat:whn3b6totfggtl7z6ph5iouw5u

Deep 3D Portrait from a Single Image [article]

Sicheng Xu, Jiaolong Yang, Dong Chen, Fang Wen, Yu Deng, Yunde Jia, Xin Tong
2020 arXiv   pre-print
We alter pose based on the recovered geometry and apply a refinement network trained with adversarial learning to ameliorate the reprojected images and translate them to the real image domain.  ...  We represent the head geometry with a parametric 3D face model together with a depth map for other head regions including hair and ear.  ...  Head pose is determined by rotation R ∈ SO(3) and translation t ∈ R 3 and is parameterized by p ∈ R 7 with rotation represented by quaternion.  ... 
arXiv:2004.11598v1 fatcat:mvaoo4ihnrevzbrzis73hh73yy

Applying Deep Learning to Fast Radio Burst Classification

Liam Connor, Joeri van Leeuwen
2018 Astronomical Journal  
High accuracy and recall can be achieved with a labelled training set of a few thousand events.  ...  We have built training and test sets using false-positive triggers from real telescopes, along with simulated FRBs, and single pulses from pulsars.  ...  We also thank Jorn Peters, Yunfan (Gerry) Zhang, and Folkert Huizinga for useful discussions.  ... 
doi:10.3847/1538-3881/aae649 fatcat:pcq3h6e6jvg25gd7bjur3sqmkm

A trans-disciplinary review of deep learning research and its relevance for water resources scientists

Chaopeng Shen
2018 Water Resources Research  
This review paper is intended to provide water resources scientists and hydrologists in particular with a simple technical overview, trans-disciplinary progress update, and a source of inspiration about  ...  Deep learning (DL), a new-generation of artificial neural network research, has transformed industries, daily lives and various scientific disciplines in recent years.  ...  Michigan State and three anonymous reviewers whose suggestions and comments have helped to improve this paper greatly. This review paper is theoretical and does not contain any dataset to be shared.  ... 
doi:10.1029/2018wr022643 fatcat:ruopsnchg5eg5hsiccyadinf54

Label-Only Membership Inference Attacks [article]

Christopher A. Choquette-Choo, Florian Tramer, Nicholas Carlini, Nicolas Papernot
2021 arXiv   pre-print
We find that training models with differential privacy and (strong) L2 regularization are the only known defense strategies that successfully prevents all attacks.  ...  Membership inference attacks are one of the simplest forms of privacy leakage for machine learning models: given a data point and model, determine whether the point was used to train the model.  ...  standard convolu- knowledge and query access to the model. tional neural network (CNN) and a ResNet (He et al., 2015).  ... 
arXiv:2007.14321v3 fatcat:yrggxjox3faj7axb6hs6aalvdy

Disentangling Adversarial Robustness and Generalization [article]

David Stutz, Matthias Hein, Bernt Schiele
2019 arXiv   pre-print
A recent hypothesis even states that both robust and accurate models are impossible, i.e., adversarial robustness and generalization are conflicting goals.  ...  To confirm our claims, we present extensive experiments on synthetic data (with known manifold) as well as on EMNIST, Fashion-MNIST and CelebA.  ...  A rotation and a translation suffice: Fooling CNNs with simple transformations. arXiv.org, abs/1712.02779, 2017. 6 [23] Alhussein Fawzi, Omar Fawzi, and Pascal Frossard.  ... 
arXiv:1812.00740v2 fatcat:cvggtcukqra5zbvmnn4ixjqoou

Synthetic Data for Deep Learning [article]

Sergey I. Nikolenko
2019 arXiv   pre-print
In this work, we attempt to provide a comprehensive survey of the various directions in the development and application of synthetic data.  ...  adaptation at the feature/model level without explicit data transformations.  ...  They record 4 fisheye cameras with 190°horizontal field of view, a rotating LiDAR, GNSS and IMU sensors, and odometry signals with 400K frames with depth labeling and 10K frames with semantic segmentation  ... 
arXiv:1909.11512v1 fatcat:qquxnw4dfvgmfeztbpdqhr44gy

A Review on Deep Learning Techniques for Video Prediction [article]

Sergiu Oprea, Pablo Martinez-Gonzalez, Alberto Garcia-Garcia, John Alejandro Castro-Vargas, Sergio Orts-Escolano, Jose Garcia-Rodriguez, Antonis Argyros
2020 arXiv   pre-print
The summary of the datasets and methods is accompanied with experimental results that facilitate the assessment of the state of the art on a quantitative basis.  ...  The ability to predict, anticipate and reason about future outcomes is a key component of intelligent decision-making systems.  ...  [71] proposed a CNN designed to predict local affine transformations of overlapping image patches.  ... 
arXiv:2004.05214v2 fatcat:weerbkanmjb4dn6wkn5o4b5aia
« Previous Showing results 1 — 15 out of 85 results