OpenCodePapers

object-recognition-on-shape-bias

Object Recognition
Dataset Link
Results over time
Click legend items to toggle metrics. Hover points for model names.
Leaderboard
PaperCodeshape biasModelNameReleaseDate
Intriguing properties of generative classifiers✓ Link98.7Imagen2023-09-28
Intriguing properties of generative classifiers✓ Link92.7Stable Diffusion2023-09-28
Intriguing properties of generative classifiers✓ Link91.7Parti2023-09-28
Scaling Vision Transformers to 22 Billion Parameters✓ Link86.4ViT-22B-3842023-02-10
Scaling Vision Transformers to 22 Billion Parameters✓ Link83.8ViT-22B-5602023-02-10
Learning Transferable Visual Models From Natural Language Supervision✓ Link79.9CLIP (ViT-B)2021-02-26
Scaling Vision Transformers to 22 Billion Parameters✓ Link78.0ViT-22B-2242023-02-10
Do Adversarially Robust ImageNet Models Transfer Better?✓ Link69.5ResNet-50 (L2 eps 5.0 adv trained)2020-07-16
The Origins and Prevalence of Texture Bias in Convolutional Neural Networks62.2ResNet-50 (with strong augmentations)2019-11-20
Billion-scale semi-supervised learning for image classification✓ Link49.8SWSL (ResNeXt-101)2019-05-02
ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness✓ Link42.9AlexNet2018-11-29
A Simple Framework for Contrastive Learning of Visual Representations✓ Link41.7SimCLR (ResNet-50x2)2020-02-13
A Simple Framework for Contrastive Learning of Visual Representations✓ Link40.7SimCLR (ResNet-50x4)2020-02-13
A Simple Framework for Contrastive Learning of Visual Representations✓ Link38.9SimCLR (ResNet-50x1)2020-02-13
ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness✓ Link31.2GoogLeNet2018-11-29
Billion-scale semi-supervised learning for image classification✓ Link28.6SWSL (ResNet-50)2019-05-02
ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness✓ Link22.1ResNet-502018-11-29
ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness✓ Link17.2VGG-162018-11-29