A Study and Comparison of Human and Deep Learning Recognition Performance under Visual Distortions

Dodge, Samuel; Karam, Lina J.

doi:10.1109/icccn.2017.8038465

Cited by 325 publications

(242 citation statements)

References 19 publications

Supporting

Mentioning

235

Contrasting

Unclassified

Order By: Relevance

“…Regularization techniques in general are useful to build models robust to traditional perturbations [35]. Despite recent great advances, neural networks are not successful enough at dealing with corrupted images coming from real world applications [4].…”

Section: Introductionmentioning

confidence: 99%

Are Adversarial Robustness and Common Perturbation Robustness Independent Attributes ?

Laugros

Caplier

Ospici

2019

2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW)

View full text Add to dashboard Cite

Neural Networks have been shown to be sensitive to common perturbations such as blur, Gaussian noise, rotations, etc. They are also vulnerable to some artificial malicious corruptions called adversarial examples. The adversarial examples study has recently become very popular and it sometimes even reduces the term "adversarial robustness" to the term "robustness". Yet, we do not know to what extent the adversarial robustness is related to the global robustness. Similarly, we do not know if a robustness to various common perturbations such as translations or contrast losses for instance, could help with adversarial corruptions. We intend to study the links between the robustnesses of neural networks to both perturbations. With our experiments, we provide one of the first benchmark designed to estimate the robustness of neural networks to common perturbations. We show that increasing the robustness to carefully selected common perturbations, can make neural networks more robust to unseen common perturbations. We also prove that adversarial robustness and robustness to common perturbations are independent. Our results make us believe that neural network robustness should be addressed in a broader sense.

show abstract

Section: Introductionmentioning

confidence: 99%

Are Adversarial Robustness and Common Perturbation Robustness Independent Attributes ?

Laugros

Caplier

Ospici

2019

2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW)

View full text Add to dashboard Cite

show abstract

“…where · F denotes Frobenius norm, λ is the penalty parameter, and we have assumed n l ≥ n l−1 for a certain layer l. By using increasingly larger values of λ, the problem (14) approaches to achieve strict OrthDNNs. One can use SGD based methods to solve (14), where the additional computation cost incurred by the regularizer is marginal. However, it is not straightforward to set a proper value of λ in order to strike a good balance between the two terms of (14).…”

Section: Alternative Algorithms For Approximate Orthdnnsmentioning

confidence: 99%

“…One can use SGD based methods to solve (14), where the additional computation cost incurred by the regularizer is marginal. However, it is not straightforward to set a proper value of λ in order to strike a good balance between the two terms of (14). Note that soft regularization algorithm of the type (14) is used in the related works [11], [55].…”

Section: Alternative Algorithms For Approximate Orthdnnsmentioning

confidence: 99%

See 1 more Smart Citation

Orthogonal Deep Neural Networks

Jia

Wen

et al. 2021

IEEE Trans. Pattern Anal. Mach. Intell.

View full text Add to dashboard Cite

In this paper, we introduce the algorithms of Orthogonal Deep Neural Networks (OrthDNNs) to connect with recent interest of spectrally regularized deep learning methods. OrthDNNs are theoretically motivated by generalization analysis of modern DNNs, with the aim to find solution properties of network weights that guarantee better generalization. To this end, we first prove that DNNs are of local isometry on data distributions of practical interest; by using a new covering of the sample space and introducing the local isometry property of DNNs into generalization analysis, we establish a new generalization error bound that is both scale-and range-sensitive to singular value spectrum of each of networks' weight matrices. We prove that the optimal bound w.r.t. the degree of isometry is attained when each weight matrix has a spectrum of equal singular values, among which orthogonal weight matrix or a non-square one with orthonormal rows or columns is the most straightforward choice, suggesting the algorithms of OrthDNNs. We present both algorithms of strict and approximate OrthDNNs, and for the later ones we propose a simple yet effective algorithm called Singular Value Bounding (SVB), which performs as well as strict OrthDNNs, but at a much lower computational cost. We also propose Bounded Batch Normalization (BBN) to make compatible use of batch normalization with OrthDNNs. We conduct extensive comparative studies by using modern architectures on benchmark image classification. Experiments show the efficacy of OrthDNNs.

show abstract

“…While machine learning systems might have surpassed humans at many tasks [9], they generally need far more data to reach the same level of performance. Nonetheless, it is not completely fair to compare humans to algorithms directly, since humans enter a task with a large amount of prior knowledge.…”

Section: Related Workmentioning

confidence: 99%

Semi Few-Shot Attribute Translation

Durall

Pfreundt

Keuper

2019

2019 International Conference on Image and Vision Computing New Zealand (IVCNZ)

View full text Add to dashboard Cite

Recent studies have shown remarkable success in imageto-image translation for attribute transfer applications. However, most of existing approaches are based on deep learning and require an abundant amount of labeled data to produce good results, therefore limiting their applicability. In the same vein, recent advances in meta-learning have led to successful implementations with limited available data, allowing so-called few-shot learning.In this paper, we address this limitation of supervised methods, by proposing a novel approach based on GANs. These are trained in a meta-training manner, which allows them to perform image-to-image translations using just a few labeled samples from a new target class. This work empirically demonstrates the potential of training a GAN for few shot image-to-image translation on hair color attribute synthesis tasks, opening the door to further research on generative transfer learning.

show abstract

A Study and Comparison of Human and Deep Learning Recognition Performance under Visual Distortions

Cited by 325 publications

References 19 publications

Are Adversarial Robustness and Common Perturbation Robustness Independent Attributes ?

Are Adversarial Robustness and Common Perturbation Robustness Independent Attributes ?

Orthogonal Deep Neural Networks

Semi Few-Shot Attribute Translation

Contact Info

Product

Resources

About