# Image Classification on ImageNet (val) (Intermediate Feature Metrics)

Source: [Image Classification on ImageNet (val) (Intermediate Feature Metrics)](<https://www.sota2.com/research/sota/image-classification-on-imagenet-val-intermediate-feature-metrics>)

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

Data updated: 2026-02-26T05:45:13.592Z

Evaluation of Image Classification performance on the validation split of the ImageNet dataset, specifically measured using intermediate feature metrics (conv2, conv3, conv4, conv5).

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 | ImageNet Classification |
| Dataset (including any stated split) | ImageNet (val) |
| Metric | Feature Score (conv2) |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 8 of 8 |
| 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

- [Feature Score (conv2)](<https://www.sota2.com/research/sota/image-classification-on-imagenet-val-intermediate-feature-metrics.md?metric=Feature%20Score%20(conv2)>): Higher is better; unit: SCALAR.

- [Feature Score (conv3)](<https://www.sota2.com/research/sota/image-classification-on-imagenet-val-intermediate-feature-metrics.md?metric=Feature%20Score%20(conv3)>): Higher is better; unit: SCALAR.

- [Feature Score (conv4)](<https://www.sota2.com/research/sota/image-classification-on-imagenet-val-intermediate-feature-metrics.md?metric=Feature%20Score%20(conv4)>): Higher is better; unit: SCALAR.

- [Feature Score (conv5)](<https://www.sota2.com/research/sota/image-classification-on-imagenet-val-intermediate-feature-metrics.md?metric=Feature%20Score%20(conv5)>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Feature Score (conv2) | Result date | Report paper |
| --- | --- | --- | --- |
| Krizhevsky et al. | 56.5 | Not recorded | 68fa62d41e209131ce9fb63e |
| Noroozi & Favaro | 56 | Not recorded | 68fa62d41e209131ce9fb63e |
| Donahue et al. | 51.9 | Not recorded | 68fa62d41e209131ce9fb63e |
| Krizhevsky et al. | 50.5 | Not recorded | 68fa62d41e209131ce9fb63e |
| Doersch et al. | 47.6 | Not recorded | 68fa62d41e209131ce9fb63e |
| Wang & Gupta | 46.9 | Not recorded | 68fa62d41e209131ce9fb63e |
| Colorful Image Colorization | 46.6 | Not recorded | 68fa62d41e209131ce9fb63e |
| Gaussian | 41 | Not recorded | 68fa62d41e209131ce9fb63e |

### Krizhevsky et al.

Feature Score (conv2): 56.5. Result ID: 698be279ecace47ef5599141.

[Reporting paper](<https://www.sota2.com/research/paper/68fa62d41e209131ce9fb63e.md>)

Recorded evaluation configuration: {"Training":"labels","Input":"rgb","Params":"1.00","Feats":"1.00","Runtime":"1.00"}. This record does not assert independent reproduction.

### Noroozi & Favaro

Feature Score (conv2): 56. Result ID: 698be279ecace47ef5599143.

[Reporting paper](<https://www.sota2.com/research/paper/68fa62d41e209131ce9fb63e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa639f1e209131ce9ffc68.md>)

Recorded evaluation configuration: {"Training":"imagenet","Input":"rgb","Params":"1.00","Feats":"5.60","Runtime":"7.35"}. This record does not assert independent reproduction.

### Donahue et al.

Feature Score (conv2): 51.9. Result ID: 698be279ecace47ef5599147.

[Reporting paper](<https://www.sota2.com/research/paper/68fa62d41e209131ce9fb63e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa633c1e209131ce9fdb2d.md>)

Recorded evaluation configuration: {"Training":"imagenet","Input":"rgb","Params":"1.00","Feats":"0.87","Runtime":"0.96"}. This record does not assert independent reproduction.

### Krizhevsky et al.

Feature Score (conv2): 50.5. Result ID: 698be279ecace47ef5599142.

[Reporting paper](<https://www.sota2.com/research/paper/68fa62d41e209131ce9fb63e.md>)

Recorded evaluation configuration: {"Training":"labels","Input":"L","Params":"0.99","Feats":"1.00","Runtime":"0.92"}. This record does not assert independent reproduction.

### Doersch et al.

Feature Score (conv2): 47.6. Result ID: 698be279ecace47ef5599145.

[Reporting paper](<https://www.sota2.com/research/paper/68fa62d41e209131ce9fb63e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa62411e209131ce9f8348.md>)

Recorded evaluation configuration: {"Training":"imagenet","Input":"rgb","Params":"1.61","Feats":"1.00","Runtime":"2.82"}. This record does not assert independent reproduction.

### Wang & Gupta

Feature Score (conv2): 46.9. Result ID: 698be279ecace47ef5599146.

[Reporting paper](<https://www.sota2.com/research/paper/68fa62d41e209131ce9fb63e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa62171e209131ce9f7793.md>)

Recorded evaluation configuration: {"Training":"videos","Input":"rgb","Params":"1.00","Feats":"1.00","Runtime":"1.00"}. This record does not assert independent reproduction.

### Colorful Image Colorization

Feature Score (conv2): 46.6. Result ID: 698be279ecace47ef5599148.

[Reporting paper](<https://www.sota2.com/research/paper/68fa62d41e209131ce9fb63e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa62d41e209131ce9fb63e.md>)

[Implementation](<http://richzhang.github.io/colorization/>)

Recorded evaluation configuration: {"Training":"imagenet","Input":"L","Params":"0.99","Feats":"0.87","Runtime":"0.84"}. This record does not assert independent reproduction.

### Gaussian

Feature Score (conv2): 41. Result ID: 698be279ecace47ef5599144.

[Reporting paper](<https://www.sota2.com/research/paper/68fa62d41e209131ce9fb63e.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa639f1e209131ce9ffc68.md>)

Recorded evaluation configuration: {"Training":"imagenet","Input":"rgb","Params":"1.00","Feats":"1.00","Runtime":"1.00"}. This record does not assert independent reproduction.

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