# Image Classification on ImageNet DEEPNETS-1M (test)

Source: [Image Classification on ImageNet DEEPNETS-1M (test)](<https://www.sota2.com/research/sota/image-classification-on-imagenet-deepnets-1m-test>)

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

Data updated: 2026-02-26T05:36:26.773Z

Evaluation of Image Classification performance on the test split of the ImageNet DEEPNETS-1M dataset, reporting average and maximum Top-5 accuracy.

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 DEEPNETS-1M (test) |
| Metric | Avg Top-5 Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 5 of 5 |
| 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

- [Avg Top-5 Accuracy](<https://www.sota2.com/research/sota/image-classification-on-imagenet-deepnets-1m-test.md?metric=Avg%20Top-5%20Accuracy>): Higher is better; unit: PERCENT.

- [Max Top-5 Accuracy](<https://www.sota2.com/research/sota/image-classification-on-imagenet-deepnets-1m-test.md?metric=Max%20Top-5%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Avg Top-5 Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| SGD (10000 steps) | 37.7 | Not recorded | 68fa6b1b1e209131cea2c408 |
| GHN-2 | 27.2 | Not recorded | 68fa6b1b1e209131cea2c408 |
| SGD (5000 steps) | 25.6 | Not recorded | 68fa6b1b1e209131cea2c408 |
| GHN-1 | 17.2 | Not recorded | 68fa6b1b1e209131cea2c408 |
| SGD (1 step) | 0.5 | Not recorded | 68fa6b1b1e209131cea2c408 |

### SGD (10000 steps)

Avg Top-5 Accuracy: 37.7. Result ID: 698e5dbeffc3901de09cfb52.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b1b1e209131cea2c408.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa63671e209131ce9fe75b.md>)

Recorded evaluation configuration: {"Number of updates":"10k","GPU compute time (sec)":"4e3","CPU compute time (sec)":"6e4"}. This record does not assert independent reproduction.

### GHN-2

Avg Top-5 Accuracy: 27.2. Result ID: 698e5dbeffc3901de09cfb4f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b1b1e209131cea2c408.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b1b1e209131cea2c408.md>)

[Implementation](<https://github.com/facebookresearch/ppuda>)

Recorded evaluation configuration: {"Number of updates":"1","GPU compute time (sec)":"0.3","CPU compute time (sec)":"0.7"}. This record does not assert independent reproduction.

### SGD (5000 steps)

Avg Top-5 Accuracy: 25.6. Result ID: 698e5dbeffc3901de09cfb51.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b1b1e209131cea2c408.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa63671e209131ce9fe75b.md>)

Recorded evaluation configuration: {"Number of updates":"5k","GPU compute time (sec)":"2e3","CPU compute time (sec)":"3e4"}. This record does not assert independent reproduction.

### GHN-1

Avg Top-5 Accuracy: 17.2. Result ID: 698e5dbeffc3901de09cfb4e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b1b1e209131cea2c408.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa68f01e209131cea1f17b.md>)

Recorded evaluation configuration: {"Number of updates":"1","GPU compute time (sec)":"0.3","CPU compute time (sec)":"0.5"}. This record does not assert independent reproduction.

### SGD (1 step)

Avg Top-5 Accuracy: 0.5. Result ID: 698e5dbeffc3901de09cfb50.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6b1b1e209131cea2c408.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa63671e209131ce9fe75b.md>)

Recorded evaluation configuration: {"Number of updates":"1","GPU compute time (sec)":"0.4","CPU compute time (sec)":"6.0"}. This record does not assert independent reproduction.

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