ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness

R Geirhos, P Rubisch, C Michaelis, M Bethge… - arXiv preprint arXiv …, 2018 - arxiv.org
Convolutional Neural Networks (CNNs) are commonly thought to recognise objects by
learning increasingly complex representations of object shapes. Some recent studies …
被引用次数：2795 相关文章

ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness

R Geirhos, P Rubisch, C Michaelis… - International …, 2018 - openreview.net
Convolutional Neural Networks (CNNs) are commonly thought to recognise objects by
learning increasingly complex representations of object shapes. Some recent studies …

[引用][C] ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness

R Geirhos, P Rubisch, C Michaelis, M Bethge… - arXiv, 2018 - ub01.uni-tuebingen.de
ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy
and robustness ImageNet-trained CNNs are biased towards texture; increasing shape bias …

[引用][C] ImageNet-trained CNNs are biased toward texture; increasing shape bias improves accuracy and robustness

G Robert - Retrieved from https://arXiv: 1811.12231, 2018 - cir.nii.ac.jp
ImageNet-trained CNNs are biased toward texture; increasing shape bias improves accuracy
and robustness | CiNii Research CiNii 国立情報学研究所 学術情報ナビゲータ[サイニィ] 詳細へ …
[PDF] 180.76.120.163

[PDF][PDF] IMAGENET-TRAINED CNNS ARE BIASED TOWARDS TEXTURE; INCREASING SHAPE BIAS IMPROVES ACCURACY AND ROBUSTNESS

R Geirhos, P Rubisch, C Michaelis, M Bethge… - 180.76.120.163
ABSTRACT Convolutional Neural Networks (CNNs) are commonly thought to recognise
objects by learning increasingly complex representations of object shapes. Some recent …

ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness

R Geirhos, P Rubisch, C Michaelis, M Bethge… - arXiv e …, 2018 - ui.adsabs.harvard.edu
Abstract Convolutional Neural Networks (CNNs) are commonly thought to recognise objects
by learning increasingly complex representations of object shapes. Some recent studies …
[PDF] utdallas.edu

[PDF][PDF] IMAGENET-TRAINED CNNS ARE BIASED TOWARDS TEXTURE; INCREASING SHAPE BIAS IMPROVES ACCURACY AND ROBUSTNESS

R Geirhos, P Rubisch, C Michaelis, M Bethge… - utdallas.edu
ABSTRACT Convolutional Neural Networks (CNNs) are commonly thought to recognise
objects by learning increasingly complex representations of object shapes. Some recent …

[引用][C] ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness

R Geirhos, P Rubisch, C Michaelis, M Bethge… - …, 2018 - tobias-lib.ub.uni-tuebingen.de
ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy
and robustness ImageNet-trained CNNs are biased towards texture; increasing shape bias …
[PDF] openreview.net

[PDF][PDF] IMAGENET-TRAINED CNNS ARE BIASED TOWARDS TEXTURE; INCREASING SHAPE BIAS IMPROVES ACCURACY AND ROBUSTNESS

R Geirhos, P Rubisch, C Michaelis, M Bethge… - openreview.net
ABSTRACT Convolutional Neural Networks (CNNs) are commonly thought to recognise
objects by learning increasingly complex representations of object shapes. Some recent …
[PDF] utdallas.edu

[PDF][PDF] IMAGENET-TRAINED CNNS ARE BIASED TOWARDS TEXTURE; INCREASING SHAPE BIAS IMPROVES ACCURACY AND ROBUSTNESS

R Geirhos, P Rubisch, C Michaelis, M Bethge… - utdallas.edu
ABSTRACT Convolutional Neural Networks (CNNs) are commonly thought to recognise
objects by learning increasingly complex representations of object shapes. Some recent …