Object Recognition on shape bias

Metric: shape bias (higher is better)

LeaderboardDataset

Loading chart...

Results

Sort:

#	Model↕	shape bias▼	Extra Data	Paper	Date↕	Code
1	Imagen	98.7	No	Intriguing properties of generative classifiers	2023-09-28	Code
2	Stable Diffusion	92.7	No	Intriguing properties of generative classifiers	2023-09-28	Code
3	Parti	91.7	No	Intriguing properties of generative classifiers	2023-09-28	Code
4	ViT-22B-384	86.4	No	Scaling Vision Transformers to 22 Billion Parame...	2023-02-10	Code
5	ViT-22B-560	83.8	No	Scaling Vision Transformers to 22 Billion Parame...	2023-02-10	Code
6	CLIP (ViT-B)	79.9	No	Learning Transferable Visual Models From Natural...	2021-02-26	Code
7	ViT-22B-224	78	No	Scaling Vision Transformers to 22 Billion Parame...	2023-02-10	Code
8	ResNet-50 (L2 eps 5.0 adv trained)	69.5	No	Do Adversarially Robust ImageNet Models Transfer...	2020-07-16	Code
9	ResNet-50 (with strong augmentations)	62.2	No	The Origins and Prevalence of Texture Bias in Co...	2019-11-20	-
10	SWSL (ResNeXt-101)	49.8	No	Billion-scale semi-supervised learning for image...	2019-05-02	Code
11	AlexNet	42.9	No	ImageNet-trained CNNs are biased towards texture...	2018-11-29	Code
12	SimCLR (ResNet-50x2)	41.7	No	A Simple Framework for Contrastive Learning of V...	2020-02-13	Code
13	SimCLR (ResNet-50x4)	40.7	No	A Simple Framework for Contrastive Learning of V...	2020-02-13	Code
14	SimCLR (ResNet-50x1)	38.9	No	A Simple Framework for Contrastive Learning of V...	2020-02-13	Code
15	GoogLeNet	31.2	No	ImageNet-trained CNNs are biased towards texture...	2018-11-29	Code
16	SWSL (ResNet-50)	28.6	No	Billion-scale semi-supervised learning for image...	2019-05-02	Code
17	ResNet-50	22.1	No	ImageNet-trained CNNs are biased towards texture...	2018-11-29	Code
18	VGG-16	17.2	No	ImageNet-trained CNNs are biased towards texture...	2018-11-29	Code

#1ImagenSOTA
98.7
shape bias· 2023-09-28
Intriguing properties of generative classifiers Code
#2Stable Diffusion
92.7
shape bias· 2023-09-28
Intriguing properties of generative classifiers Code
#3Parti
91.7
shape bias· 2023-09-28
Intriguing properties of generative classifiers Code
#4ViT-22B-384SOTA
86.4
shape bias· 2023-02-10
Scaling Vision Transformers to 22 Billion Parameters Code
#5ViT-22B-560
83.8
shape bias· 2023-02-10
Scaling Vision Transformers to 22 Billion Parameters Code
#6CLIP (ViT-B)SOTA
79.9
shape bias· 2021-02-26
Learning Transferable Visual Models From Natural Language Supervision Code
#7ViT-22B-224
78
shape bias· 2023-02-10
Scaling Vision Transformers to 22 Billion Parameters Code
#8ResNet-50 (L2 eps 5.0 adv trained)SOTA
69.5
shape bias· 2020-07-16
Do Adversarially Robust ImageNet Models Transfer Better?Code
#9ResNet-50 (with strong augmentations)SOTA
62.2
shape bias· 2019-11-20
The Origins and Prevalence of Texture Bias in Convolutional Neural Networks
#10SWSL (ResNeXt-101)SOTA
49.8
shape bias· 2019-05-02
Billion-scale semi-supervised learning for image classification Code
#11AlexNetSOTA
42.9
shape bias· 2018-11-29
ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness Code
#12SimCLR (ResNet-50x2)
41.7
shape bias· 2020-02-13
A Simple Framework for Contrastive Learning of Visual Representations Code
#13SimCLR (ResNet-50x4)
40.7
shape bias· 2020-02-13
A Simple Framework for Contrastive Learning of Visual Representations Code
#14SimCLR (ResNet-50x1)
38.9
shape bias· 2020-02-13
A Simple Framework for Contrastive Learning of Visual Representations Code
#15GoogLeNet
31.2
shape bias· 2018-11-29
ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness Code
#16SWSL (ResNet-50)
28.6
shape bias· 2019-05-02
Billion-scale semi-supervised learning for image classification Code
#17ResNet-50
22.1
shape bias· 2018-11-29
ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness Code
#18VGG-16
17.2
shape bias· 2018-11-29
ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness Code