| VOLO: Vision Outlooker for Visual Recognition | ✓ Link | 57.2 | 51.8 | 59.7 | VOLO-D5 | 2021-06-24 |
| A ConvNet for the 2020s | ✓ Link | 53.5 | 46.9 | 56 | ConvNeXt-B | 2022-01-10 |
| Aggregated Residual Transformations for Deep Neural Networks | ✓ Link | 51.7 | 48.1 | 54.8 | ResNeXt-101 32x16d | 2016-11-16 |
| Adversarial Examples Improve Image Recognition | ✓ Link | 50.5 | 45.8 | 53.2 | EfficientNet-B8 (advprop+autoaug) | 2019-11-21 |
| Adversarial Examples Improve Image Recognition | ✓ Link | 49.7 | 45 | 52 | EfficientNet-B7 (advprop+autoaug) | 2019-11-21 |
| Adversarial Examples Improve Image Recognition | ✓ Link | 49.6 | 44.7 | 53.2 | EfficientNet-B6 (advprop+autoaug) | 2019-11-21 |
| Adversarial Examples Improve Image Recognition | ✓ Link | 49.1 | 44 | 51.7 | EfficientNet-B5 (advprop+autoaug) | 2019-11-21 |
| An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale | ✓ Link | 49 | | | ViT-16/L-224 | 2020-10-22 |
| ResNet strikes back: An improved training procedure in timm | ✓ Link | 48.9 | 39.1 | 44.4 | ResNet-50 (gn) | 2021-10-01 |
| Adversarial Examples Improve Image Recognition | ✓ Link | 48.1 | 42.5 | 51.4 | EfficientNet-B4 (advprop+autoaug) | 2019-11-21 |
| Deep Residual Learning for Image Recognition | ✓ Link | 47.5 | 43.3 | 51.3 | ResNet-152 | 2015-12-10 |
| Deep Residual Learning for Image Recognition | ✓ Link | 46.3 | 40.5 | 50.1 | ResNet-101 | 2015-12-10 |
| AutoAugment: Learning Augmentation Strategies From Data | ✓ Link | 45.8 | 39.3 | 50.7 | EfficientNet-B6 (autoaug) | 2019-06-01 |
| AutoAugment: Learning Augmentation Strategies From Data | ✓ Link | 45.7 | 39.8 | 50.2 | EfficientNet-B5 (autoaug) | 2019-06-01 |
| Adversarial Examples Improve Image Recognition | ✓ Link | 45.5 | 39.8 | 49.5 | EfficientNet-B3 (advprop+autoaug) | 2019-11-21 |
| AutoAugment: Learning Augmentation Strategies From Data | ✓ Link | 45 | 39.1 | 49.9 | EfficientNet-B7 (autoaug) | 2019-06-01 |
| RandAugment: Practical automated data augmentation with a reduced search space | ✓ Link | 45 | 38.9 | 48.7 | EfficientNet-B7 (randaug) | 2019-09-30 |
| AutoAugment: Learning Augmentation Strategies From Data | ✓ Link | 44.3 | 38.2 | 48.6 | EfficientNet-B4 (autoaug) | 2019-06-01 |
| Adversarial Examples Improve Image Recognition | ✓ Link | 44.3 | 38.2 | 48 | EfficientNet-B2 (advprop+autoaug) | 2019-11-21 |
| Deep Residual Learning for Image Recognition | ✓ Link | 42.9 | 37.1 | 47.7 | ResNet-50 | 2015-12-10 |
| EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks | ✓ Link | 42.8 | 37 | 47.3 | EfficientNet-B5 | 2019-05-28 |
| AutoAugment: Learning Augmentation Policies from Data | ✓ Link | 42.6 | 34.9 | 47.5 | EfficientNet-B3 (autoaug) | 2018-05-24 |
| Adversarial Examples Improve Image Recognition | ✓ Link | 42.4 | 36.2 | 46.7 | EfficientNet-B1 (advprop+autoaug) | 2019-11-21 |
| AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty | ✓ Link | 42.2 | 35.9 | 46.4 | ResNet-50 (augmix) | 2019-12-05 |
| RandAugment: Practical automated data augmentation with a reduced search space | ✓ Link | 42.1 | 35.5 | 47.3 | EfficientNet-B5 (randaug) | 2019-09-30 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 41.7 | 35.7 | 46.1 | ResNet-101 (lpf3) | 2019-04-25 |
| EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks | ✓ Link | 41.7 | 35.6 | 46.4 | EfficientNet-B4 | 2019-05-28 |
| AutoAugment: Learning Augmentation Policies from Data | ✓ Link | 41.6 | 34.3 | 45.8 | EfficientNet-B2 (autoaug) | 2018-05-24 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 41.5 | 35.2 | 45.3 | ResNet-50 (lpf5) | 2019-04-25 |
| The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalization | ✓ Link | 41.3 | 34.9 | 46 | ResNet-50 (deepaugment) | 2020-06-29 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 41.1 | 35.1 | 45.2 | ResNet-101 (lpf2) | 2019-04-25 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 41 | 34.8 | 45.8 | ResNet-101 (lpf5) | 2019-04-25 |
| EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks | ✓ Link | 40.7 | 34.2 | 45.3 | EfficientNet-B3 | 2019-05-28 |
| Adversarial Examples Improve Image Recognition | ✓ Link | 40.5 | 34.2 | 44.9 | EfficientNet-B0 (advprop+autoaug) | 2019-11-21 |
| The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalization | ✓ Link | 40.3 | 34.1 | 44.5 | ResNet-50 (deepaugment+augmix) | 2020-06-29 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 40.3 | 33.4 | 45.1 | ResNet-50 (lpf2) | 2019-04-25 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 40 | 34.3 | 44.7 | ResNet-50 (lpf3) | 2019-04-25 |
| Bag of Tricks for Image Classification with Convolutional Neural Networks | ✓ Link | 39.7 | 35.8 | 43.5 | ResNet-26-D | 2018-12-04 |
| AutoAugment: Learning Augmentation Policies from Data | ✓ Link | 39.7 | 32.8 | 44.4 | EfficientNet-B1 (autoaug) | 2018-05-24 |
| ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness | ✓ Link | 39.2 | 32.4 | 44.6 | ResNet-50 (SIN_IN_IN) | 2018-11-29 |
| Measuring Robustness to Natural Distribution Shifts in Image Classification | ✓ Link | 38.8 | 33.6 | 42.9 | ResNet-50 (IN-C) | 2020-07-01 |
| Measuring Robustness to Natural Distribution Shifts in Image Classification | ✓ Link | 38.8 | 32.5 | 43.5 | ResNet-50 (IN-C_brightness) | 2020-07-01 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 38.7 | 32 | 42.7 | DenseNet121 (lpf5) | 2019-04-25 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 38.3 | 32.4 | 42.8 | ResNet-34 (lpf2) | 2019-04-25 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 38.3 | 32.3 | 42.8 | DenseNet-121 (lpf3) | 2019-04-25 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 38.3 | 31.9 | 42.9 | ResNet-34 (lpf3) | 2019-04-25 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 38.3 | 31.7 | 43.1 | DenseNet-121 (lpf2) | 2019-04-25 |
| Measuring Robustness to Natural Distribution Shifts in Image Classification | ✓ Link | 38.3 | 31.4 | 42.7 | ResNet-50 (IN-C_spatter) | 2020-07-01 |
| ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness | ✓ Link | 38.2 | 32.5 | 42.7 | ResNet-50 (SIN_IN) | 2018-11-29 |
| Measuring Robustness to Natural Distribution Shifts in Image Classification | ✓ Link | 38.2 | 32.4 | 42.4 | ResNet-50 (IN-C_saturate) | 2020-07-01 |
| EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks | ✓ Link | 38.1 | 31.4 | 42.8 | EfficientNet-B2 | 2019-05-28 |
| Measuring Robustness to Natural Distribution Shifts in Image Classification | ✓ Link | 37.4 | 30.9 | 41.4 | ResNet-50 (IN-C_pixelate) | 2020-07-01 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 37.2 | 31.3 | 41.8 | VGG-16 BN (lpf2) | 2019-04-25 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 37.2 | 29.9 | 42.5 | ResNet-34 (lpf5) | 2019-04-25 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 37 | 30.8 | 41.7 | VGG-16 BN (lpf5) | 2019-04-25 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 36.9 | 30.6 | 42.1 | VGG-16 BN (lpf3) | 2019-04-25 |
| Very Deep Convolutional Networks for Large-Scale Image Recognition | ✓ Link | 36.7 | 31.1 | 41.1 | VGG-16 BN | 2014-09-04 |
| EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks | ✓ Link | 36.7 | 30.9 | 41.5 | EfficientNet-B1 | 2019-05-28 |
| Measuring Robustness to Natural Distribution Shifts in Image Classification | ✓ Link | 36.5 | 30.7 | 40.9 | ResNet-50 (IN-C_contrast) | 2020-07-01 |
| Measuring Robustness to Natural Distribution Shifts in Image Classification | ✓ Link | 36.5 | 30.3 | 41.3 | ResNet-50 (IN-C_jpeg_compression) | 2020-07-01 |
| Measuring Robustness to Natural Distribution Shifts in Image Classification | ✓ Link | 36.4 | 30.2 | 40.6 | ResNet-50 (IN-C_gaussian_noise) | 2020-07-01 |
| Very Deep Convolutional Networks for Large-Scale Image Recognition | ✓ Link | 36.2 | 29.4 | 40.8 | VGG-19 BN | 2014-09-04 |
| Measuring Robustness to Natural Distribution Shifts in Image Classification | ✓ Link | 36.1 | 29.7 | 40 | ResNet-50 (IN-C_frost) | 2020-07-01 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 36 | 30.4 | 40.3 | MobileNetV2 (lpf3) | 2019-04-25 |
| Measuring Robustness to Natural Distribution Shifts in Image Classification | ✓ Link | 35.9 | 30.3 | 39.9 | ResNet-50 (IN-C_fog_aws) | 2020-07-01 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 35.8 | 29.1 | 40.1 | MobileNetV2 (lpf5) | 2019-04-25 |
| Measuring Robustness to Natural Distribution Shifts in Image Classification | ✓ Link | 35.7 | 30.2 | 39.6 | ResNet-50 (IN-C_motion_blur) | 2020-07-01 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 35.6 | 28.5 | 39.5 | ResNet-18 (lpf3) | 2019-04-25 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 35.5 | 30.3 | 39.2 | MobileNetV2 (lpf2) | 2019-04-25 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 35.5 | 28.7 | 40.1 | ResNet-18 (lpf2) | 2019-04-25 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 35.1 | 28.2 | 40 | VGG-16 (lpf3) | 2019-04-25 |
| AutoAugment: Learning Augmentation Policies from Data | ✓ Link | 34.9 | 27.3 | 40.1 | EfficientNet-B0 (autoaug) | 2018-05-24 |
| Very Deep Convolutional Networks for Large-Scale Image Recognition | ✓ Link | 34.7 | 29 | 39.3 | VGG-19 | 2014-09-04 |
| Very Deep Convolutional Networks for Large-Scale Image Recognition | ✓ Link | 34.7 | 28.5 | 39.5 | VGG-16 | 2014-09-04 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 34.7 | 27.7 | 38.9 | ResNet-18 (lpf5) | 2019-04-25 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 34.5 | 27.8 | 39.4 | VGG-16 (lpf5) | 2019-04-25 |
| EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks | ✓ Link | 34.2 | 27.4 | 38.4 | EfficientNet-B0 | 2019-05-28 |
| Very Deep Convolutional Networks for Large-Scale Image Recognition | ✓ Link | 33.7 | 28.3 | 38.4 | VGG-13 BN | 2014-09-04 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 33.5 | 26.7 | 38.5 | VGG-16 (lpf2) | 2019-04-25 |
| Very Deep Convolutional Networks for Large-Scale Image Recognition | ✓ Link | 32.9 | 25.8 | 37.1 | VGG-11 BN | 2014-09-04 |
| Measuring Robustness to Natural Distribution Shifts in Image Classification | ✓ Link | 32.7 | 28.3 | 36.6 | ResNet-50 (IN-C_zoom_blur) | 2020-07-01 |
| Very Deep Convolutional Networks for Large-Scale Image Recognition | ✓ Link | 32.4 | 26.4 | 36.5 | VGG-13 | 2014-09-04 |
| Very Deep Convolutional Networks for Large-Scale Image Recognition | ✓ Link | 31.5 | 25.2 | 36.1 | VGG-11 | 2014-09-04 |
| Measuring Robustness to Natural Distribution Shifts in Image Classification | ✓ Link | 30.2 | 24.3 | 34.3 | ResNet-50 (IN-C_greyscale) | 2020-07-01 |
| Adversarial Training for Free! | ✓ Link | 26.7 | 20.5 | 30.9 | ResNet-50 (adv-train-free) | 2019-04-29 |
| ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness | ✓ Link | 25.3 | 20.4 | 30 | ResNet-50 (SIN) | 2018-11-29 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 23.1 | 17.5 | 26.8 | AlexNet (lpf3) | 2019-04-25 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 22.8 | 18.2 | 26.8 | AlexNet (lpf2) | 2019-04-25 |
| Making Convolutional Networks Shift-Invariant Again | ✓ Link | 22.7 | 18.4 | 26.8 | AlexNet (lpf5) | 2019-04-25 |
| An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale | ✓ Link | | | 450 | ViT-8/B-224 | 2020-10-22 |