ML Research Wiki / Benchmarks / Object Recognition / shape bias

shape bias

Object Recognition Benchmark

Performance Over Time

📊 Showing 18 results | 📏 Metric: shape bias

Top Performing Models

Rank	Model	Paper	shape bias	Date	Code
1	Imagen	Intriguing properties of generative classifiers	98.70	2023-09-28	📦 SamsungSAILMontreal/ForestDiffusion
2	Stable Diffusion	Intriguing properties of generative classifiers	92.70	2023-09-28	📦 SamsungSAILMontreal/ForestDiffusion
3	Parti	Intriguing properties of generative classifiers	91.70	2023-09-28	📦 SamsungSAILMontreal/ForestDiffusion
4	ViT-22B-384	Scaling Vision Transformers to 22 Billion Parameters	86.40	2023-02-10	📦 lucidrains/flash-cosine-sim-attention
5	ViT-22B-560	Scaling Vision Transformers to 22 Billion Parameters	83.80	2023-02-10	📦 lucidrains/flash-cosine-sim-attention
6	CLIP (ViT-B)	Learning Transferable Visual Models From Natural Language Supervision	79.90	2021-02-26	📦 openai/CLIP 📦 mlfoundations/open_clip 📦 towhee-io/towhee
7	ViT-22B-224	Scaling Vision Transformers to 22 Billion Parameters	78.00	2023-02-10	📦 lucidrains/flash-cosine-sim-attention
8	ResNet-50 (L2 eps 5.0 adv trained)	Do Adversarially Robust ImageNet Models Transfer Better?	69.50	2020-07-16	📦 MadryLab/robustness 📦 lengstrom/gitlinks
9	ResNet-50 (with strong augmentations)	The Origins and Prevalence of Texture Bias in Convolutional Neural Networks	62.20	2019-11-20	-
10	SWSL (ResNeXt-101)	Billion-scale semi-supervised learning for image classification	49.80	2019-05-02	📦 facebookresearch/semi-supervised-ImageNet1K-models 📦 leaderj1001/Billion-scale-semi-supervised-learning 📦 salesforce/ensemble-of-averages 📦 tiskw/patchcore-ad

All Papers (18)

Intriguing properties of generative classifiers

2023

Imagen

SamsungSAILMontreal/ForestDiffusion

Intriguing properties of generative classifiers

2023

Stable Diffusion

SamsungSAILMontreal/ForestDiffusion

Intriguing properties of generative classifiers

2023

Parti

SamsungSAILMontreal/ForestDiffusion

Scaling Vision Transformers to 22 Billion Parameters

2023

ViT-22B-384

lucidrains/flash-cosine-sim-attention

Scaling Vision Transformers to 22 Billion Parameters

2023

ViT-22B-560

lucidrains/flash-cosine-sim-attention

Learning Transferable Visual Models From Natural Language Supervision

2021

CLIP (ViT-B)

openai/CLIP mlfoundations/open_clip

Scaling Vision Transformers to 22 Billion Parameters

2023

ViT-22B-224

lucidrains/flash-cosine-sim-attention

Do Adversarially Robust ImageNet Models Transfer Better?

2020

ResNet-50 (L2 eps 5.0 adv trained)

MadryLab/robustness lengstrom/gitlinks

The Origins and Prevalence of Texture Bias in Convolutional Neural Networks

2019

ResNet-50 (with strong augmentations)

Billion-scale semi-supervised learning for image classification

2019

SWSL (ResNeXt-101)

facebookresearch/semi-supervised-ImageNet1K-models leaderj1001/Billion-scale-semi-supervised-learning

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

2018

AlexNet

rgeirhos/texture-vs-shape rgeirhos/Stylized-ImageNet

A Simple Framework for Contrastive Learning of Visual Representations

2020

SimCLR (ResNet-50x2)

tensorflow/models google-research/simclr

A Simple Framework for Contrastive Learning of Visual Representations

2020

SimCLR (ResNet-50x4)

tensorflow/models google-research/simclr

A Simple Framework for Contrastive Learning of Visual Representations

2020

SimCLR (ResNet-50x1)

tensorflow/models google-research/simclr

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

2018

GoogLeNet

rgeirhos/texture-vs-shape rgeirhos/Stylized-ImageNet

Billion-scale semi-supervised learning for image classification

2019

SWSL (ResNet-50)

facebookresearch/semi-supervised-ImageNet1K-models leaderj1001/Billion-scale-semi-supervised-learning

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

2018

ResNet-50

rgeirhos/texture-vs-shape rgeirhos/Stylized-ImageNet

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

2018

VGG-16

rgeirhos/texture-vs-shape rgeirhos/Stylized-ImageNet

shape bias

Performance Over Time

Edit Benchmark Results

Edit Result

Top Performing Models

All Papers (18)

Intriguing properties of generative classifiers

Intriguing properties of generative classifiers

Intriguing properties of generative classifiers

Scaling Vision Transformers to 22 Billion Parameters

Scaling Vision Transformers to 22 Billion Parameters

Learning Transferable Visual Models From Natural Language Supervision

Scaling Vision Transformers to 22 Billion Parameters

Do Adversarially Robust ImageNet Models Transfer Better?

The Origins and Prevalence of Texture Bias in Convolutional Neural Networks

Billion-scale semi-supervised learning for image classification

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

A Simple Framework for Contrastive Learning of Visual Representations

A Simple Framework for Contrastive Learning of Visual Representations

A Simple Framework for Contrastive Learning of Visual Representations

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

Billion-scale semi-supervised learning for image classification

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

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

Model	Paper	shape bias	Date
Imagen	Intriguing properties of generative classifiers	98.70	2023-09-28
Stable Diffusion	Intriguing properties of generative classifiers	92.70	2023-09-28
Parti	Intriguing properties of generative classifiers	91.70	2023-09-28
ViT-22B-384	Scaling Vision Transformers to 22 Billion Paramet…	86.40	2023-02-10
ViT-22B-560	Scaling Vision Transformers to 22 Billion Paramet…	83.80	2023-02-10
CLIP (ViT-B)	Learning Transferable Visual Models From Natural …	79.90	2021-02-26
ViT-22B-224	Scaling Vision Transformers to 22 Billion Paramet…	78.00	2023-02-10
ResNet-50 (L2 eps 5.0 adv trained)	Do Adversarially Robust ImageNet Models Transfer …	69.50	2020-07-16
ResNet-50 (with strong augmentations)	The Origins and Prevalence of Texture Bias in Con…	62.20	2019-11-20
SWSL (ResNeXt-101)	Billion-scale semi-supervised learning for image …	49.80	2019-05-02
AlexNet	ImageNet-trained CNNs are biased towards texture;…	42.90	2018-11-29
SimCLR (ResNet-50x2)	A Simple Framework for Contrastive Learning of Vi…	41.70	2020-02-13
SimCLR (ResNet-50x4)	A Simple Framework for Contrastive Learning of Vi…	40.70	2020-02-13
SimCLR (ResNet-50x1)	A Simple Framework for Contrastive Learning of Vi…	38.90	2020-02-13
GoogLeNet	ImageNet-trained CNNs are biased towards texture;…	31.20	2018-11-29
SWSL (ResNet-50)	Billion-scale semi-supervised learning for image …	28.60	2019-05-02
ResNet-50	ImageNet-trained CNNs are biased towards texture;…	22.10	2018-11-29
VGG-16	ImageNet-trained CNNs are biased towards texture;…	17.20	2018-11-29