# Image Classification on TinyImageNet (test) (Corruption Metrics)

Source: [Image Classification on TinyImageNet (test) (Corruption Metrics)](<https://www.sota2.com/research/sota/image-classification-on-tinyimagenet-test-corruption-metrics>)

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

Data updated: 2026-02-26T05:31:50.426Z

Evaluation of Image Classification performance on the test split of the TinyImageNet dataset, focusing on accuracy under various corruption levels.

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) | TinyImageNet (test) |
| Metric | Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 16 of 16 |
| 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

- [Accuracy](<https://www.sota2.com/research/sota/image-classification-on-tinyimagenet-test-corruption-metrics.md?metric=Accuracy>): Higher is better; unit: PERCENT.

- [Corruption Metric C](<https://www.sota2.com/research/sota/image-classification-on-tinyimagenet-test-corruption-metrics.md?metric=Corruption%20Metric%20C>): Higher is better; unit: PERCENT.

- [C15 Score](<https://www.sota2.com/research/sota/image-classification-on-tinyimagenet-test-corruption-metrics.md?metric=C15%20Score>): Higher is better; unit: PERCENT.

- [Mean Corruption Accuracy](<https://www.sota2.com/research/sota/image-classification-on-tinyimagenet-test-corruption-metrics.md?metric=Mean%20Corruption%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| SYN+NST+TA (Ours) | 74.69 | Not recorded | 69438aa5423ea422657295ea |
| Synthetic | 73.37 | Not recorded | 69438aa5423ea422657295ea |
| SYN+NST+TA | 73.07 | Not recorded | 69438aa5423ea422657295ea |
| SYN+NST+TA + Random Erasing + Noise + Feature Noise | 72.43 | Not recorded | 69438aa5423ea422657295ea |
| SYN+NST+TA + Random Erasing + Noise | 71.8 | Not recorded | 69438aa5423ea422657295ea |
| Noisy Feature Mixup + TrivialAugment | 69.93 | Not recorded | 69438aa5423ea422657295ea |
| TrivialAugment + Random Erasing + Noise | 69.23 | Not recorded | 69438aa5423ea422657295ea |
| Mixup + Cutmix | 68.26 | Not recorded | 69438aa5423ea422657295ea |
| TrivialAugment | 68.01 | Not recorded | 69438aa5423ea422657295ea |
| NoisyMix | 67.83 | Not recorded | 69438aa5423ea422657295ea |
| NoisyMix | 67.82 | Not recorded | 69438aa5423ea422657295ea |
| Random Erasing | 64.56 | Not recorded | 69438aa5423ea422657295ea |
| Random Erasing + Noise | 63.88 | Not recorded | 69438aa5423ea422657295ea |
| AugMix | 62.36 | Not recorded | 69438aa5423ea422657295ea |
| AugMix + DeepAugment | 62.07 | Not recorded | 69438aa5423ea422657295ea |
| Baseline | 61.95 | Not recorded | 69438aa5423ea422657295ea |

### SYN+NST+TA (Ours)

Accuracy: 74.69. Result ID: 699835af5698665ec4914feb.

[Reporting paper](<https://www.sota2.com/research/paper/69438aa5423ea422657295ea.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69438aa5423ea422657295ea.md>)

Recorded evaluation configuration: {"Backbone":"WRN-28-4","Training Epochs":"600"}. This record does not assert independent reproduction.

### Synthetic

Accuracy: 73.37. Result ID: 699835af5698665ec4915005.

[Reporting paper](<https://www.sota2.com/research/paper/69438aa5423ea422657295ea.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6da51e209131cea3b271.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### SYN+NST+TA

Accuracy: 73.07. Result ID: 699835af5698665ec4915006.

[Reporting paper](<https://www.sota2.com/research/paper/69438aa5423ea422657295ea.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69438aa5423ea422657295ea.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### SYN+NST+TA + Random Erasing + Noise + Feature Noise

Accuracy: 72.43. Result ID: 699835af5698665ec4915008.

[Reporting paper](<https://www.sota2.com/research/paper/69438aa5423ea422657295ea.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69438aa5423ea422657295ea.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### SYN+NST+TA + Random Erasing + Noise

Accuracy: 71.8. Result ID: 699835af5698665ec4915007.

[Reporting paper](<https://www.sota2.com/research/paper/69438aa5423ea422657295ea.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69438aa5423ea422657295ea.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### Noisy Feature Mixup + TrivialAugment

Accuracy: 69.93. Result ID: 699835af5698665ec4915003.

[Reporting paper](<https://www.sota2.com/research/paper/69438aa5423ea422657295ea.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### TrivialAugment + Random Erasing + Noise

Accuracy: 69.23. Result ID: 699835af5698665ec4914fff.

[Reporting paper](<https://www.sota2.com/research/paper/69438aa5423ea422657295ea.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### Mixup + Cutmix

Accuracy: 68.26. Result ID: 699835af5698665ec4915002.

[Reporting paper](<https://www.sota2.com/research/paper/69438aa5423ea422657295ea.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### TrivialAugment

Accuracy: 68.01. Result ID: 699835af5698665ec4914ffc.

[Reporting paper](<https://www.sota2.com/research/paper/69438aa5423ea422657295ea.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a8c1e209131cea28b5b.md>)

[Implementation](<https://github.com/automl/trivialaugment>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### NoisyMix

Accuracy: 67.83. Result ID: 699835af5698665ec4915004.

[Reporting paper](<https://www.sota2.com/research/paper/69438aa5423ea422657295ea.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c971e209131cea34af3.md>)

Recorded evaluation configuration: {"Training Strategy":"300 epoch rerun without early stopping"}. This record does not assert independent reproduction.

### NoisyMix

Accuracy: 67.82. Result ID: 699835af5698665ec4914fea.

[Reporting paper](<https://www.sota2.com/research/paper/69438aa5423ea422657295ea.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c971e209131cea34af3.md>)

Recorded evaluation configuration: {"Backbone":"WRN-28-4","Training Epochs":"600"}. This record does not assert independent reproduction.

### Random Erasing

Accuracy: 64.56. Result ID: 699835af5698665ec4914ffd.

[Reporting paper](<https://www.sota2.com/research/paper/69438aa5423ea422657295ea.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa63d41e209131cea010d6.md>)

[Implementation](<https://github.com/zhunzhong07/Random-Erasing>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### Random Erasing + Noise

Accuracy: 63.88. Result ID: 699835af5698665ec4914ffe.

[Reporting paper](<https://www.sota2.com/research/paper/69438aa5423ea422657295ea.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### AugMix

Accuracy: 62.36. Result ID: 699835af5698665ec4915000.

[Reporting paper](<https://www.sota2.com/research/paper/69438aa5423ea422657295ea.md>)

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

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

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### AugMix + DeepAugment

Accuracy: 62.07. Result ID: 699835af5698665ec4915001.

[Reporting paper](<https://www.sota2.com/research/paper/69438aa5423ea422657295ea.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### Baseline

Accuracy: 61.95. Result ID: 699835af5698665ec4914ffb.

[Reporting paper](<https://www.sota2.com/research/paper/69438aa5423ea422657295ea.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

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