# Image Classification on ImageNet and ImageNet-V2

Source: [Image Classification on ImageNet and ImageNet-V2](<https://www.sota2.com/research/sota/image-classification-on-imagenet-and-imagenet-v2>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-02-26T05:47:01.594Z

Evaluation of Image Classification performance across both the original ImageNet dataset and the ImageNet-V2 dataset, typically reporting separate scores and an overall average.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Image Classification |
| Dataset (including any stated split) | ImageNet and ImageNet-V2 |
| Metric | ImageNet Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 17 of 17 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [ImageNet Accuracy](<https://www.sota2.com/research/sota/image-classification-on-imagenet-and-imagenet-v2.md?metric=ImageNet%20Accuracy>): Higher is better; unit: PERCENT.

- [ImageNet-V2 Accuracy](<https://www.sota2.com/research/sota/image-classification-on-imagenet-and-imagenet-v2.md?metric=ImageNet-V2%20Accuracy>): Higher is better; unit: PERCENT.

- [Average Accuracy](<https://www.sota2.com/research/sota/image-classification-on-imagenet-and-imagenet-v2.md?metric=Average%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | ImageNet Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| AugReg-ViT + DAT + DAD | 81.9 | Not recorded | 68fa721c1e209131cea548b7 |
| AugReg-ViT + Pyramid AT | 81.7 | Not recorded | 68fa721c1e209131cea548b7 |
| AugReg-ViT + DAT | 81.5 | Not recorded | 68fa721c1e209131cea548b7 |
| CLIP | 79.9 | Not recorded | 68fa721c1e209131cea548b7 |
| AugReg-ViT | 79.9 | Not recorded | 68fa721c1e209131cea548b7 |
| AugReg-ViT + DAD | 79.6 | Not recorded | 68fa721c1e209131cea548b7 |
| Advprop | 79.5 | Not recorded | 68fa721c1e209131cea548b7 |
| Debiased | 79.3 | Not recorded | 68fa721c1e209131cea548b7 |
| Fast Advprop | 79 | Not recorded | 68fa721c1e209131cea548b7 |
| DAT + DAD | 77.8 | Not recorded | 68fa721c1e209131cea548b7 |
| Advprop | 77.6 | Not recorded | 68fa721c1e209131cea548b7 |
| Debiased | 76.9 | Not recorded | 68fa721c1e209131cea548b7 |
| DAT | 76.5 | Not recorded | 68fa721c1e209131cea548b7 |
| ResNet50 | 76.1 | Not recorded | 68fa721c1e209131cea548b7 |
| DAD | 75.7 | Not recorded | 68fa721c1e209131cea548b7 |
| Pyramid AT | 75.5 | Not recorded | 68fa721c1e209131cea548b7 |
| ViT | 72.8 | Not recorded | 68fa721c1e209131cea548b7 |

### AugReg-ViT + DAT + DAD

ImageNet Accuracy: 81.9. Result ID: 699205343a6c3148490c18b7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b7.md>)

[Implementation](<https://github.com/andyz245/DiscreteAdversarialDistillation>)

Recorded evaluation configuration: {"Backbone":"ViT-B/16","Resolution":"224x224"}. This record does not assert independent reproduction.

### AugReg-ViT + Pyramid AT

ImageNet Accuracy: 81.7. Result ID: 699205343a6c3148490c18b4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d5a1e209131cea3965e.md>)

[Implementation](<https://www.sota2.com/pyramidat.github.io>)

Recorded evaluation configuration: {"Backbone":"ViT-B/16","Resolution":"224x224"}. This record does not assert independent reproduction.

### AugReg-ViT + DAT

ImageNet Accuracy: 81.5. Result ID: 699205343a6c3148490c18b5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d6f1e209131cea39d91.md>)

[Implementation](<https://github.com/alibaba/easyrobust>)

Recorded evaluation configuration: {"Backbone":"ViT-B/16","Resolution":"224x224"}. This record does not assert independent reproduction.

### CLIP

ImageNet Accuracy: 79.9. Result ID: 699205343a6c3148490c18ae.

[Reporting paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69581e209131cea21ad0.md>)

[Implementation](<https://github.com/OpenAI/CLIP>)

Recorded evaluation configuration: {"Backbone":"ViT-L/14","Resolution":"224x224"}. This record does not assert independent reproduction.

### AugReg-ViT

ImageNet Accuracy: 79.9. Result ID: 699205343a6c3148490c18b3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6cd51e209131cea36117.md>)

Recorded evaluation configuration: {"Backbone":"ViT-B/16","Resolution":"224x224"}. This record does not assert independent reproduction.

### AugReg-ViT + DAD

ImageNet Accuracy: 79.6. Result ID: 699205343a6c3148490c18b6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b7.md>)

[Implementation](<https://github.com/andyz245/DiscreteAdversarialDistillation>)

Recorded evaluation configuration: {"Backbone":"ViT-B/16","Resolution":"224x224"}. This record does not assert independent reproduction.

### Advprop

ImageNet Accuracy: 79.5. Result ID: 699205343a6c3148490c18b0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa675c1e209131cea15dee.md>)

[Implementation](<https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet>)

Recorded evaluation configuration: {"Backbone":"ViT-B/16","Resolution":"224x224"}. This record does not assert independent reproduction.

### Debiased

ImageNet Accuracy: 79.3. Result ID: 699205343a6c3148490c18b2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa699a1e209131cea23030.md>)

[Implementation](<https://github.com/LiYingwei/ShapeTextureDebiasedTraining>)

Recorded evaluation configuration: {"Backbone":"ViT-B/16","Resolution":"224x224"}. This record does not assert independent reproduction.

### Fast Advprop

ImageNet Accuracy: 79. Result ID: 699205343a6c3148490c18b1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c551e209131cea3359a.md>)

Recorded evaluation configuration: {"Backbone":"ViT-B/16","Resolution":"224x224"}. This record does not assert independent reproduction.

### DAT + DAD

ImageNet Accuracy: 77.8. Result ID: 699205343a6c3148490c18be.

[Reporting paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b7.md>)

[Implementation](<https://github.com/andyz245/DiscreteAdversarialDistillation>)

Recorded evaluation configuration: {"Backbone":"ResNet50","Resolution":"224x224"}. This record does not assert independent reproduction.

### Advprop

ImageNet Accuracy: 77.6. Result ID: 699205343a6c3148490c18b9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa675c1e209131cea15dee.md>)

[Implementation](<https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet>)

Recorded evaluation configuration: {"Backbone":"ResNet50","Resolution":"224x224"}. This record does not assert independent reproduction.

### Debiased

ImageNet Accuracy: 76.9. Result ID: 699205343a6c3148490c18bb.

[Reporting paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa699a1e209131cea23030.md>)

[Implementation](<https://github.com/LiYingwei/ShapeTextureDebiasedTraining>)

Recorded evaluation configuration: {"Backbone":"ResNet50","Resolution":"224x224"}. This record does not assert independent reproduction.

### DAT

ImageNet Accuracy: 76.5. Result ID: 699205343a6c3148490c18bc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d6f1e209131cea39d91.md>)

[Implementation](<https://github.com/alibaba/easyrobust>)

Recorded evaluation configuration: {"Backbone":"ResNet50","Resolution":"224x224"}. This record does not assert independent reproduction.

### ResNet50

ImageNet Accuracy: 76.1. Result ID: 699205343a6c3148490c18b8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa62321e209131ce9f8328.md>)

Recorded evaluation configuration: {"Backbone":"ResNet50","Resolution":"224x224"}. This record does not assert independent reproduction.

### DAD

ImageNet Accuracy: 75.7. Result ID: 699205343a6c3148490c18bd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b7.md>)

[Implementation](<https://github.com/andyz245/DiscreteAdversarialDistillation>)

Recorded evaluation configuration: {"Backbone":"ResNet50","Resolution":"224x224"}. This record does not assert independent reproduction.

### Pyramid AT

ImageNet Accuracy: 75.5. Result ID: 699205343a6c3148490c18ba.

[Reporting paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6d5a1e209131cea3965e.md>)

[Implementation](<https://www.sota2.com/pyramidat.github.io>)

Recorded evaluation configuration: {"Backbone":"ResNet50","Resolution":"224x224"}. This record does not assert independent reproduction.

### ViT

ImageNet Accuracy: 72.8. Result ID: 699205343a6c3148490c18af.

[Reporting paper](<https://www.sota2.com/research/paper/68fa721c1e209131cea548b7.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a0b1e209131cea25b04.md>)

[Implementation](<https://github.com/google-research/vision_transformer>)

Recorded evaluation configuration: {"Backbone":"ViT-B/16","Resolution":"224x224"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.