# Image Classification on TinyImageNet (NPPR/AR Robustness)

Source: [Image Classification on TinyImageNet (NPPR/AR Robustness)](<https://www.sota2.com/research/sota/image-classification-on-tinyimagenet-nppr-ar-robustness>)

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

Data updated: 2026-06-01T09:24:51.362Z

Evaluation of Image Classification performance on the TinyImageNet dataset using specialized metrics including NPPR, PR (Gaussian/Uniform/Laplace), and Adversarial Robustness (AR) scores against PGD, CW, and AA attacks.

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 |
| Metric | Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 4 of 4 |
| 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-nppr-ar-robustness.md?metric=Accuracy>): Higher is better; unit: PERCENT.

- [NPPR (Ours)](<https://www.sota2.com/research/sota/image-classification-on-tinyimagenet-nppr-ar-robustness.md?metric=NPPR%20(Ours)>): Higher is better; unit: PERCENT.

- [PR (Gaussian)](<https://www.sota2.com/research/sota/image-classification-on-tinyimagenet-nppr-ar-robustness.md?metric=PR%20(Gaussian)>): Higher is better; unit: PERCENT.

- [PR (Uniform)](<https://www.sota2.com/research/sota/image-classification-on-tinyimagenet-nppr-ar-robustness.md?metric=PR%20(Uniform)>): Higher is better; unit: PERCENT.

- [PR (Laplace)](<https://www.sota2.com/research/sota/image-classification-on-tinyimagenet-nppr-ar-robustness.md?metric=PR%20(Laplace)>): Higher is better; unit: PERCENT.

- [AR (PGD)](<https://www.sota2.com/research/sota/image-classification-on-tinyimagenet-nppr-ar-robustness.md?metric=AR%20(PGD)>): Higher is better; unit: SCALAR.

- [AR (CW)](<https://www.sota2.com/research/sota/image-classification-on-tinyimagenet-nppr-ar-robustness.md?metric=AR%20(CW)>): Higher is better; unit: SCALAR.

- [AR (AA)](<https://www.sota2.com/research/sota/image-classification-on-tinyimagenet-nppr-ar-robustness.md?metric=AR%20(AA)>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| WRN50 | 73.74 | Not recorded | 6923e677d6b0dd334292eef2 |
| ResNet50 | 70.39 | Not recorded | 6923e677d6b0dd334292eef2 |
| VGG16 | 66.95 | Not recorded | 6923e677d6b0dd334292eef2 |
| ResNet18 | 56.99 | Not recorded | 6923e677d6b0dd334292eef2 |

### WRN50

Accuracy: 73.74. Result ID: 6a1d4edefbc373592c16cdf9.

[Reporting paper](<https://www.sota2.com/research/paper/6923e677d6b0dd334292eef2.md>)

Recorded evaluation configuration: {"Joint dependency modes":"7","L\_inf perturbation radius":"16/255"}. This record does not assert independent reproduction.

### ResNet50

Accuracy: 70.39. Result ID: 6a1d4edefbc373592c16cdf7.

[Reporting paper](<https://www.sota2.com/research/paper/6923e677d6b0dd334292eef2.md>)

Recorded evaluation configuration: {"Joint dependency modes":"7","L\_inf perturbation radius":"16/255"}. This record does not assert independent reproduction.

### VGG16

Accuracy: 66.95. Result ID: 6a1d4edefbc373592c16cdf8.

[Reporting paper](<https://www.sota2.com/research/paper/6923e677d6b0dd334292eef2.md>)

Recorded evaluation configuration: {"Joint dependency modes":"7","L\_inf perturbation radius":"16/255"}. This record does not assert independent reproduction.

### ResNet18

Accuracy: 56.99. Result ID: 6a1d4edefbc373592c16cdf6.

[Reporting paper](<https://www.sota2.com/research/paper/6923e677d6b0dd334292eef2.md>)

Recorded evaluation configuration: {"Joint dependency modes":"7","L\_inf perturbation radius":"16/255"}. This record does not assert independent reproduction.

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