# Semantic Segmentation on PASCAL VOC without background category 2012

Source: [Semantic Segmentation on PASCAL VOC without background category 2012](<https://www.sota2.com/research/sota/semantic-segmentation-on-pascal-voc-without-background-category-2012>)

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

Data updated: 2026-02-26T06:10:33.794Z

Evaluation of standard Semantic Segmentation performance on the PASCAL VOC 2012 dataset, specifically excluding the background category from evaluation.

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 | Semantic Segmentation |
| Dataset (including any stated split) | PASCAL VOC without background category 2012 |
| Metric | mIoU |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 9 of 9 |
| 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

- [mIoU](<https://www.sota2.com/research/sota/semantic-segmentation-on-pascal-voc-without-background-category-2012.md?metric=mIoU>): Higher is better; unit: PERCENT.

## Recorded results

| Method | mIoU | Result date | Report paper |
| --- | --- | --- | --- |
| SCLIP | 80.4 | Not recorded | 68fa77a01e209131cea74770 |
| COSMOS w/ SCLIP | 77.7 | Not recorded | 68fa77a01e209131cea74770 |
| SILC-C | 77.5 | Not recorded | 68fa77a01e209131cea74770 |
| MaskCLIP | 74.9 | Not recorded | 68fa77a01e209131cea74770 |
| COSMOS | 53.6 | Not recorded | 68fa77a01e209131cea74770 |
| OpenCLIP | 51.3 | Not recorded | 68fa77a01e209131cea74770 |
| OpenCLIP | 51.3 | Not recorded | 68fa77a01e209131cea74770 |
| CLIP | 49.4 | Not recorded | 68fa77a01e209131cea74770 |
| OpenCLIP | 43.8 | Not recorded | 68fa77a01e209131cea74770 |

### SCLIP

mIoU: 80.4. Result ID: 699556f33badfaffe647a1fe.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77a01e209131cea74770.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa756e1e209131cea67d7c.md>)

Recorded evaluation configuration: {"Data Size":"400M","Zero-shot":true,"Vision Encoder Architecture":"ViT-B/16"}. This record does not assert independent reproduction.

### COSMOS w/ SCLIP

mIoU: 77.7. Result ID: 699556f33badfaffe647a200.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77a01e209131cea74770.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77a01e209131cea74770.md>)

[Implementation](<https://www.sota2.com/ExplainableML/cosmos>)

Recorded evaluation configuration: {"Data Size":"30M","Zero-shot":true,"Vision Encoder Architecture":"ViT-B/16"}. This record does not assert independent reproduction.

### SILC-C

mIoU: 77.5. Result ID: 699556f33badfaffe647a1ff.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77a01e209131cea74770.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71641e209131cea511d4.md>)

Recorded evaluation configuration: {"Data Size":"10B","Zero-shot":true,"Vision Encoder Architecture":"ViT-B/16"}. This record does not assert independent reproduction.

### MaskCLIP

mIoU: 74.9. Result ID: 699556f33badfaffe647a1fd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77a01e209131cea74770.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6f0a1e209131cea437fe.md>)

[Implementation](<https://github.com/LightDXY/MaskCLIP>)

Recorded evaluation configuration: {"Data Size":"400M","Zero-shot":true,"Vision Encoder Architecture":"ViT-B/16"}. This record does not assert independent reproduction.

### COSMOS

mIoU: 53.6. Result ID: 699556f33badfaffe647a1fb.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77a01e209131cea74770.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77a01e209131cea74770.md>)

[Implementation](<https://www.sota2.com/ExplainableML/cosmos>)

Recorded evaluation configuration: {"Data Size":"30M","Zero-shot":true,"Vision Encoder Architecture":"ViT-B/16"}. This record does not assert independent reproduction.

### OpenCLIP

mIoU: 51.3. Result ID: 699556f33badfaffe647a1f9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77a01e209131cea74770.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74261e209131cea6003e.md>)

[Implementation](<https://github.com/LAION-AI/scaling-laws-openclip>)

Recorded evaluation configuration: {"Data Size":"1B","Zero-shot":true,"Vision Encoder Architecture":"ViT-B/16"}. This record does not assert independent reproduction.

### OpenCLIP

mIoU: 51.3. Result ID: 699556f33badfaffe647a1fa.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77a01e209131cea74770.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74261e209131cea6003e.md>)

[Implementation](<https://github.com/LAION-AI/scaling-laws-openclip>)

Recorded evaluation configuration: {"Data Size":"2B","Zero-shot":true,"Vision Encoder Architecture":"ViT-B/16"}. This record does not assert independent reproduction.

### CLIP

mIoU: 49.4. Result ID: 699556f33badfaffe647a1fc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77a01e209131cea74770.md>)

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

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

Recorded evaluation configuration: {"Data Size":"400M","Zero-shot":true,"Vision Encoder Architecture":"ViT-B/16"}. This record does not assert independent reproduction.

### OpenCLIP

mIoU: 43.8. Result ID: 699556f33badfaffe647a1f8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77a01e209131cea74770.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74261e209131cea6003e.md>)

[Implementation](<https://github.com/LAION-AI/scaling-laws-openclip>)

Recorded evaluation configuration: {"Data Size":"400M","Zero-shot":true,"Vision Encoder Architecture":"ViT-B/16"}. This record does not assert independent reproduction.

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