# Generalization to Novel Objects on OPRA (test)

Source: [Generalization to Novel Objects on OPRA (test)](<https://www.sota2.com/research/sota/generalization-to-novel-objects-on-opra-test>)

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

Data updated: 2026-02-26T05:47:46.072Z

Evaluation of model generalization capabilities on novel objects using the test split of the OPRA dataset, measured by KLD, SIM, AUC-J, and NSS.

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 | Generalization to Novel Objects |
| Dataset (including any stated split) | OPRA (test) |
| Metric | KLD |
| Unit | SCALAR |
| Direction | Lower 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

- [KLD](<https://www.sota2.com/research/sota/generalization-to-novel-objects-on-opra-test.md?metric=KLD>): Lower is better; unit: SCALAR.

- [SIM](<https://www.sota2.com/research/sota/generalization-to-novel-objects-on-opra-test.md?metric=SIM>): Higher is better; unit: SCALAR.

- [AUC-J](<https://www.sota2.com/research/sota/generalization-to-novel-objects-on-opra-test.md?metric=AUC-J>): Higher is better; unit: PERCENT.

- [NSS](<https://www.sota2.com/research/sota/generalization-to-novel-objects-on-opra-test.md?metric=NSS>): Higher is better; unit: PERCENT.

## Recorded results

| Method | KLD | Result date | Report paper |
| --- | --- | --- | --- |
| HAG-Net | 1.366 | Not recorded | 68fa6a921e209131cea28f27 |
| Hotspot | 1.381 | Not recorded | 68fa6a921e209131cea28f27 |
| Img2heatmap | 1.431 | Not recorded | 68fa6a921e209131cea28f27 |
| SALGAN | 1.698 | Not recorded | 68fa6a921e209131cea28f27 |
| DeepGaze II | 1.757 | Not recorded | 68fa6a921e209131cea28f27 |
| Egogaze | 2.083 | Not recorded | 68fa6a921e209131cea28f27 |
| MLNet | 2.458 | Not recorded | 68fa6a921e209131cea28f27 |
| Center bias | 6.281 | Not recorded | 68fa6a921e209131cea28f27 |

### HAG-Net

KLD: 1.366. Result ID: 699abe715698665ec44c8ca0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a921e209131cea28f27.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a921e209131cea28f27.md>)

Recorded evaluation configuration: {"Supervision Level":"Weakly supervised"}. This record does not assert independent reproduction.

### Hotspot

KLD: 1.381. Result ID: 699abe715698665ec44c8c9f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a921e209131cea28f27.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa658f1e209131cea0b2d9.md>)

[Implementation](<http://vision.cs.utexas.edu/projects/interaction-hotspots/>)

Recorded evaluation configuration: {"Supervision Level":"Weakly supervised"}. This record does not assert independent reproduction.

### Img2heatmap

KLD: 1.431. Result ID: 699abe715698665ec44c8ca1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a921e209131cea28f27.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa658f1e209131cea0b2d9.md>)

[Implementation](<http://vision.cs.utexas.edu/projects/interaction-hotspots/>)

Recorded evaluation configuration: {"Supervision Level":"Strongly supervised"}. This record does not assert independent reproduction.

### SALGAN

KLD: 1.698. Result ID: 699abe715698665ec44c8c9e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a921e209131cea28f27.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa64871e209131cea053b2.md>)

Recorded evaluation configuration: {"Supervision Level":"Weakly supervised"}. This record does not assert independent reproduction.

### DeepGaze II

KLD: 1.757. Result ID: 699abe715698665ec44c8c9d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a921e209131cea28f27.md>)

Recorded evaluation configuration: {"Supervision Level":"Weakly supervised"}. This record does not assert independent reproduction.

### Egogaze

KLD: 2.083. Result ID: 699abe715698665ec44c8c9b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a921e209131cea28f27.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa651c1e209131cea085fe.md>)

Recorded evaluation configuration: {"Supervision Level":"Weakly supervised"}. This record does not assert independent reproduction.

### MLNet

KLD: 2.458. Result ID: 699abe715698665ec44c8c9c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a921e209131cea28f27.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa63861e209131ce9ff3e0.md>)

Recorded evaluation configuration: {"Supervision Level":"Weakly supervised"}. This record does not assert independent reproduction.

### Center bias

KLD: 6.281. Result ID: 699abe715698665ec44c8c9a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa6a921e209131cea28f27.md>)

Recorded evaluation configuration: {"Supervision Level":"Baseline"}. This record does not assert independent reproduction.

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