# Shadow detection on 3D physical scene properties (test)

Source: [Shadow detection on 3D physical scene properties (test)](<https://www.sota2.com/research/sota/shadow-detection-on-3d-physical-scene-properties-test>)

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

Data updated: 2026-02-26T06:00:22.026Z

Evaluation of Shadow detection performance on the test split of the 3D physical scene properties dataset, measured using ROC AUC.

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 | Shadow detection |
| Dataset (including any stated split) | 3D physical scene properties (test) |
| Metric | ROC AUC |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 6 of 6 |
| 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

- [ROC AUC](<https://www.sota2.com/research/sota/shadow-detection-on-3d-physical-scene-properties-test.md?metric=ROC%20AUC>): Higher is better; unit: PERCENT.

## Recorded results

| Method | ROC AUC | Result date | Report paper |
| --- | --- | --- | --- |
| DINOv2 | 95.8 | Not recorded | 68fa73851e209131cea5ce68 |
| Stable Diffusion | 95.4 | Not recorded | 68fa73851e209131cea5ce68 |
| OpenCLIP | 94.9 | Not recorded | 68fa73851e209131cea5ce68 |
| DINOv1 | 92.2 | Not recorded | 68fa73851e209131cea5ce68 |
| VQGAN | 89 | Not recorded | 68fa73851e209131cea5ce68 |
| Random | 50 | Not recorded | 68fa73851e209131cea5ce68 |

### DINOv2

ROC AUC: 95.8. Result ID: 6992e52e3a6c314849123921.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73851e209131cea5ce68.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72181e209131cea546c3.md>)

[Implementation](<https://github.com/facebookresearch/dinov2>)

Recorded evaluation configuration: {"Backbone":"ViT-G","Pre-training dataset":"LVD-142M","Evaluation protocol":"SVM grid search (best layer and C)"}. This record does not assert independent reproduction.

### Stable Diffusion

ROC AUC: 95.4. Result ID: 6992e52e3a6c314849123923.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73851e209131cea5ce68.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c391e209131cea32e4a.md>)

[Implementation](<https://github.com/CompVis/latent-diffusion>)

Recorded evaluation configuration: {"Evaluation protocol":"SVM grid search (best time step, layer and C)"}. This record does not assert independent reproduction.

### OpenCLIP

ROC AUC: 94.9. Result ID: 6992e52e3a6c31484912391f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73851e209131cea5ce68.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69581e209131cea21ad0.md>)

[Implementation](<https://github.com/OpenAI/CLIP>)

Recorded evaluation configuration: {"Backbone":"ViT-G","Pre-training dataset":"LAION","Evaluation protocol":"SVM grid search (best layer and C)"}. This record does not assert independent reproduction.

### DINOv1

ROC AUC: 92.2. Result ID: 6992e52e3a6c314849123920.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73851e209131cea5ce68.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a071e209131cea2594c.md>)

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

Recorded evaluation configuration: {"Backbone":"ViT-B","Pre-training dataset":"ImageNet","Evaluation protocol":"SVM grid search (best layer and C)"}. This record does not assert independent reproduction.

### VQGAN

ROC AUC: 89. Result ID: 6992e52e3a6c314849123922.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73851e209131cea5ce68.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6a371e209131cea26a4d.md>)

[Implementation](<https://git.io/JnyvK>)

Recorded evaluation configuration: {"Backbone":"48-layer transformer","Pre-training dataset":"ImageNet","Evaluation protocol":"SVM grid search (best layer and C)"}. This record does not assert independent reproduction.

### Random

ROC AUC: 50. Result ID: 6992e52e3a6c31484912391e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa73851e209131cea5ce68.md>)

Recorded evaluation configuration: {"Description":"Random classifier"}. This record does not assert independent reproduction.

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