# Image-based Audio-Visual Segmentation Quality Assessment on MQ-RAVSBench (Unseen)

Source: [Image-based Audio-Visual Segmentation Quality Assessment on MQ-RAVSBench (Unseen)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen>)

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

Data updated: 2026-02-26T06:00:15.673Z

Evaluation of image-based audio-visual segmentation quality assessment performance using the MQ-RAVSBench (Unseen) dataset, reporting RMSE and F2 metrics across various degradation conditions.

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-based Audio-Visual Segmentation Quality Assessment |
| Dataset (including any stated split) | MQ-RAVSBench (Unseen) |
| Metric | RMSE (Perfect) |
| Unit | SCALAR |
| Direction | Lower 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

- [RMSE (Perfect)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=RMSE%20(Perfect)>): Lower is better; unit: SCALAR.

- [RMSE (Cutout H)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=RMSE%20(Cutout%20H)>): Lower is better; unit: SCALAR.

- [RMSE (Cutout M)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=RMSE%20(Cutout%20M)>): Lower is better; unit: SCALAR.

- [RMSE (Dilate H)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=RMSE%20(Dilate%20H)>): Lower is better; unit: SCALAR.

- [RMSE (Dilate M)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=RMSE%20(Dilate%20M)>): Lower is better; unit: SCALAR.

- [RMSE (Erode H)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=RMSE%20(Erode%20H)>): Lower is better; unit: SCALAR.

- [RMSE (Erode M)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=RMSE%20(Erode%20M)>): Lower is better; unit: SCALAR.

- [RMSE (Merge)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=RMSE%20(Merge)>): Lower is better; unit: SCALAR.

- [RMSE (Full\_neg)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=RMSE%20(Full_neg)>): Lower is better; unit: SCALAR.

- [RMSE (Avg)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=RMSE%20(Avg)>): Lower is better; unit: SCALAR.

- [F2-M (Perfect)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=F2-M%20(Perfect)>): Higher is better; unit: PERCENT.

- [F2-M (Cutout H)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=F2-M%20(Cutout%20H)>): Higher is better; unit: PERCENT.

- [F2-M (Cutout M)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=F2-M%20(Cutout%20M)>): Higher is better; unit: PERCENT.

- [F2-M (Dilate H)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=F2-M%20(Dilate%20H)>): Higher is better; unit: PERCENT.

- [F2-M (Dilate M)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=F2-M%20(Dilate%20M)>): Higher is better; unit: PERCENT.

- [F2-M (Erode H)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=F2-M%20(Erode%20H)>): Higher is better; unit: PERCENT.

- [F2-M (Erode M)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=F2-M%20(Erode%20M)>): Higher is better; unit: PERCENT.

- [F2-M (Merge)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=F2-M%20(Merge)>): Higher is better; unit: PERCENT.

- [F2-M (Full\_neg)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=F2-M%20(Full_neg)>): Higher is better; unit: PERCENT.

- [F2-M (Avg)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=F2-M%20(Avg)>): Higher is better; unit: PERCENT.

- [F2-A (Perfect)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=F2-A%20(Perfect)>): Higher is better; unit: PERCENT.

- [F2-A (Cutout H)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=F2-A%20(Cutout%20H)>): Higher is better; unit: PERCENT.

- [F2-A (Cutout M)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=F2-A%20(Cutout%20M)>): Higher is better; unit: PERCENT.

- [F2-A (Dilate H)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=F2-A%20(Dilate%20H)>): Higher is better; unit: PERCENT.

- [F2-A (Dilate M)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=F2-A%20(Dilate%20M)>): Higher is better; unit: PERCENT.

- [F2-A (Erode H)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=F2-A%20(Erode%20H)>): Higher is better; unit: PERCENT.

- [F2-A (Erode M)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=F2-A%20(Erode%20M)>): Higher is better; unit: PERCENT.

- [F2-A (Merge)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=F2-A%20(Merge)>): Higher is better; unit: PERCENT.

- [F2-A (Full\_neg)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=F2-A%20(Full_neg)>): Higher is better; unit: PERCENT.

- [F2-A (Avg)](<https://www.sota2.com/research/sota/image-based-audio-visual-segmentation-quality-assessment-on-mq-ravsbench-unseen.md?metric=F2-A%20(Avg)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | RMSE (Perfect) | Result date | Report paper |
| --- | --- | --- | --- |
| Video-LLaMA3-7B | 0.192 | Not recorded | 698423f3a8b01e76cc8ee63a |
| MQ-Auditor | 0.274 | Not recorded | 698423f3a8b01e76cc8ee63a |
| Gemini-3-Flash-Preview | 0.409 | Not recorded | 698423f3a8b01e76cc8ee63a |
| Qwen2.5-Omni-7B | 0.418 | Not recorded | 698423f3a8b01e76cc8ee63a |
| Ming-Flash-Omni | 0.981 | Not recorded | 698423f3a8b01e76cc8ee63a |

### Video-LLaMA3-7B

RMSE (Perfect): 0.192. Result ID: 6997513b5698665ec4a3013f.

[Reporting paper](<https://www.sota2.com/research/paper/698423f3a8b01e76cc8ee63a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79111e209131cea7bf26.md>)

Recorded evaluation configuration: {"Evaluation Protocol":"Image-based"}. This record does not assert independent reproduction.

### MQ-Auditor

RMSE (Perfect): 0.274. Result ID: 6997513b5698665ec4a30143.

[Reporting paper](<https://www.sota2.com/research/paper/698423f3a8b01e76cc8ee63a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/698423f3a8b01e76cc8ee63a.md>)

Recorded evaluation configuration: {"Evaluation Protocol":"Image-based"}. This record does not assert independent reproduction.

### Gemini-3-Flash-Preview

RMSE (Perfect): 0.409. Result ID: 6997513b5698665ec4a30142.

[Reporting paper](<https://www.sota2.com/research/paper/698423f3a8b01e76cc8ee63a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78731e209131cea78622.md>)

Recorded evaluation configuration: {"Evaluation Protocol":"Image-based"}. This record does not assert independent reproduction.

### Qwen2.5-Omni-7B

RMSE (Perfect): 0.418. Result ID: 6997513b5698665ec4a30140.

[Reporting paper](<https://www.sota2.com/research/paper/698423f3a8b01e76cc8ee63a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77a21e209131cea7482c.md>)

Recorded evaluation configuration: {"Evaluation Protocol":"Image-based"}. This record does not assert independent reproduction.

### Ming-Flash-Omni

RMSE (Perfect): 0.981. Result ID: 6997513b5698665ec4a30141.

[Reporting paper](<https://www.sota2.com/research/paper/698423f3a8b01e76cc8ee63a.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6902f0f23ef7ccd74e5e4274.md>)

Recorded evaluation configuration: {"Evaluation Protocol":"Image-based"}. This record does not assert independent reproduction.

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