Image Classification on ImageNet-200 (test)
4.1Top-1 Error RateResNeXt-101 32x8d + WSL Pretraining
Evaluation results
| Method | Links | |
|---|---|---|
| ResNeXt-101 32x8d + WSL PretrainingBackbone=ResNeXt-101 32x8d, Pre-training=WSL Pretraining (1000x data)2020.06 | 4.1 | |
| ResNeXt-101 32x4d + SEBackbone=ResNeXt-101 32x4d, Self-Attention=SE2020.06 | 5.9 | |
| ResNeXt-101 32x8d + DeepAugment + AugMixBackbone=ResNeXt-101 32x8d, Data Augmentation=DeepAugment + AugMix2020.06 | 6.1 | |
| ResNeXt-101 32x8dBackbone=ResNeXt-101 32x8d2020.06 | 6.2 | |
| ResNet-152 + SEBackbone=ResNet-152, Self-Attention=SE2020.06 | 6.6 | |
| ResNet-101 + SEBackbone=ResNet-101, Self-Attention=SE2020.06 | 6.7 | |
| ResNet-152Backbone=ResNet-1522020.06 | 6.8 | |
| ResNet-152Backbone=ResNet-1522020.06 | 6.8 | |
| ResNeXt-101 32x4dBackbone=ResNeXt-101 32x4d2020.06 | 6.8 | |
| ResNet-50 + ImageNet-21K PretrainingBackbone=ResNet-50, Pre-training=ImageNet-21K2020.06 | 7 | |
| ResNet-50 + CBAMBackbone=ResNet-50, Attention Mechanism=CBAM (Self-Attention)2020.06 | 7 | |
| ResNet-50 + ImageNet-21K PretrainingBackbone=ResNet-50, Pre-training=ImageNet-21K (10x data)2020.06 | 7 | |
| ResNet-50 + CBAMBackbone=ResNet-50, Self-Attention=CBAM2020.06 | 7 | |
| ResNet-50 + AugMixBackbone=ResNet-50, Data Augmentation=AugMix2020.06 | 7.1 | |
| ResNet-50 + AugMixBackbone=ResNet-50, Data Augmentation=AugMix2020.06 | 7.1 | |
| ResNet-101Backbone=ResNet-1012020.06 | 7.1 | |
| ResNet-50 + DeepAugmentBackbone=ResNet-50, Data Augmentation=DeepAugment2020.06 | 7.5 | |
| ResNet-50 + DeepAugmentBackbone=ResNet-50, Data Augmentation=DeepAugment2020.06 | 7.5 | |
| ResNet-50Backbone=ResNet-502020.06 | 7.9 | |
| ResNet-50Backbone=ResNet-502020.06 | 7.9 | |
| ResNet-50 + DeepAugment + AugMixBackbone=ResNet-50, Data Augmentation=DeepAugment + AugMix2020.06 | 8 | |
| ResNet-50 + DeepAugment + AugMixBackbone=ResNet-50, Data Augmentation=DeepAugment + AugMix2020.06 | 8 | |
| ResNet-50 + Speckle NoiseBackbone=ResNet-50, Data Augmentation=Speckle Noise2020.06 | 8.1 | |
| ResNet-50 + Speckle NoiseBackbone=ResNet-50, Data Augmentation=Speckle Noise2020.06 | 8.1 | |
| ResNet-50 + Style Transfer AugmentationBackbone=ResNet-50, Data Augmentation=Style Transfer Augmentation2020.06 | 8.9 | |
| ResNet-50 + Style TransferBackbone=ResNet-50, Data Augmentation=Style Transfer2020.06 | 8.9 | |
| ResNet-50 + l_inf Adversarial TrainingBackbone=ResNet-50, Training Strategy=l_inf Adversarial Training2020.06 | 25.1 | |
| ResNet-50 + L_inf Adversarial TrainingBackbone=ResNet-50, Training Strategy=L_inf Adversarial Training2020.06 | 25.1 |