# Image Classification on Real Blurry Images (test)

Source: [Image Classification on Real Blurry Images (test)](<https://www.sota2.com/research/sota/image-classification-on-real-blurry-images-test>)

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

Data updated: 2026-02-26T05:42:15.888Z

Evaluation of Image Classification performance on the test split of the Real Blurry Images dataset, typically measured using Error rate.

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) | Real Blurry Images (test) |
| Metric | Error Rate |
| Unit | PERCENT |
| Direction | Lower is better |
| Results loaded | 10 of 10 |
| 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

- [Error Rate](<https://www.sota2.com/research/sota/image-classification-on-real-blurry-images-test.md?metric=Error%20Rate>): Lower is better; unit: PERCENT.

## Recorded results

| Method | Error Rate | Result date | Report paper |
| --- | --- | --- | --- |
| ResNet-50 + DeepAugment + AugMix | 51.7 | Not recorded | 68fa6a751e209131cea28405 |
| ResNet-50 + DeepAugment | 54.2 | Not recorded | 68fa6a751e209131cea28405 |
| ResNet-152 | 54.3 | Not recorded | 68fa6a751e209131cea28405 |
| ResNet-50 + AugMix | 54.4 | Not recorded | 68fa6a751e209131cea28405 |
| ResNet-50 + ImageNet-21K Pretraining | 54.8 | Not recorded | 68fa6a751e209131cea28405 |
| ResNet-50 + CBAM | 56.5 | Not recorded | 68fa6a751e209131cea28405 |
| ResNet-50 + Style Transfer | 56.7 | Not recorded | 68fa6a751e209131cea28405 |
| ResNet-50 + Speckle Noise | 56.9 | Not recorded | 68fa6a751e209131cea28405 |
| ResNet-50 | 58.7 | Not recorded | 68fa6a751e209131cea28405 |
| ResNet-50 + l∞ Adversarial Training | 71.6 | Not recorded | 68fa6a751e209131cea28405 |

### ResNet-50 + DeepAugment + AugMix

Error Rate: 51.7. Result ID: 698e36c9ffc3901de09bb4d4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a751e209131cea28405.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a751e209131cea28405.md>)

[Implementation](<https://github.com/hendrycks/imagenet-r>)

Recorded evaluation configuration: {"Backbone":"ResNet-50","Data Augmentation":"DeepAugment + AugMix"}. This record does not assert independent reproduction.

### ResNet-50 + DeepAugment

Error Rate: 54.2. Result ID: 698e36c9ffc3901de09bb4d3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a751e209131cea28405.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a751e209131cea28405.md>)

[Implementation](<https://github.com/hendrycks/imagenet-r>)

Recorded evaluation configuration: {"Backbone":"ResNet-50","Data Augmentation":"DeepAugment"}. This record does not assert independent reproduction.

### ResNet-152

Error Rate: 54.3. Result ID: 698e36c9ffc3901de09bb4d5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a751e209131cea28405.md>)

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

Recorded evaluation configuration: {"Backbone":"ResNet-152","Model Architecture":"Larger Models"}. This record does not assert independent reproduction.

### ResNet-50 + AugMix

Error Rate: 54.4. Result ID: 698e36c9ffc3901de09bb4d2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a751e209131cea28405.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67041e209131cea13ccb.md>)

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

Recorded evaluation configuration: {"Backbone":"ResNet-50","Data Augmentation":"AugMix"}. This record does not assert independent reproduction.

### ResNet-50 + ImageNet-21K Pretraining

Error Rate: 54.8. Result ID: 698e36c9ffc3901de09bb4cd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a751e209131cea28405.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66911e209131cea10e94.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-50","Pre-training":"ImageNet-21K"}. This record does not assert independent reproduction.

### ResNet-50 + CBAM

Error Rate: 56.5. Result ID: 698e36c9ffc3901de09bb4ce.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a751e209131cea28405.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa649f1e209131cea05b3b.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-50","Model Architecture":"CBAM (Self-Attention)"}. This record does not assert independent reproduction.

### ResNet-50 + Style Transfer

Error Rate: 56.7. Result ID: 698e36c9ffc3901de09bb4d1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a751e209131cea28405.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa5f3a1e209131ce9e8680.md>)

[Implementation](<https://github.com/rgeirhos/texture-vs-shape>)

Recorded evaluation configuration: {"Backbone":"ResNet-50","Data Augmentation":"Style Transfer"}. This record does not assert independent reproduction.

### ResNet-50 + Speckle Noise

Error Rate: 56.9. Result ID: 698e36c9ffc3901de09bb4d0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a751e209131cea28405.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa680b1e209131cea19c52.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-50","Data Augmentation":"Speckle Noise"}. This record does not assert independent reproduction.

### ResNet-50

Error Rate: 58.7. Result ID: 698e36c9ffc3901de09bb4cc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a751e209131cea28405.md>)

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

Recorded evaluation configuration: {"Backbone":"ResNet-50"}. This record does not assert independent reproduction.

### ResNet-50 + l∞ Adversarial Training

Error Rate: 71.6. Result ID: 698e36c9ffc3901de09bb4cf.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a751e209131cea28405.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa66e51e209131cea12efc.md>)

Recorded evaluation configuration: {"Backbone":"ResNet-50","Adversarial Training":"l∞"}. This record does not assert independent reproduction.

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