- BiRefNet General
- Open weights · commercially permissiveLicense: MIT
- BiRefNet HRSOD
- Open weights · commercially permissiveLicense: MIT
birefnet-general vs birefnet-hrsod
Quantitative comparison: 33 images scored against ground-truth mattes. Lower is better. Use the view controls on each row to switch between transparency, chroma key, and alpha matte on all outputs.
| Metric | birefnet-general | birefnet-hrsod |
|---|---|---|
| MGE (Edge Quality) | 0.03384 ✓ | 0.04385 |
| MAE (Overall Accuracy) | 0.01691 ✓ | 0.03044 |
- MGE
- 0.03450✓
- MAE
- 0.00754✓
- MGE
- 0.07758
- MAE
- 0.03906
- MGE
- 0.03391✓
- MAE
- 0.02906✓
- MGE
- 0.04882
- MAE
- 0.09456
- MGE
- 0.03369✓
- MAE
- 0.00657✓
- MGE
- 0.04549
- MAE
- 0.01077
- MGE
- 0.01534✓
- MAE
- 0.00932✓
- MGE
- 0.02432
- MAE
- 0.01259
- MGE
- 0.01776✓
- MAE
- 0.00662✓
- MGE
- 0.02960
- MAE
- 0.01206
- MGE
- 0.05518
- MAE
- 0.02445✓
- MGE
- 0.05418✓
- MAE
- 0.02499
- MGE
- 0.02070✓
- MAE
- 0.00960
- MGE
- 0.02189
- MAE
- 0.00934✓
- MGE
- 0.06499✓
- MAE
- 0.01274✓
- MGE
- 0.08612
- MAE
- 0.01928
- MGE
- 0.01420✓
- MAE
- 0.00333✓
- MGE
- 0.02243
- MAE
- 0.00553
- MGE
- 0.03582✓
- MAE
- 0.01162✓
- MGE
- 0.04122
- MAE
- 0.01428
- MGE
- 0.02967✓
- MAE
- 0.00716✓
- MGE
- 0.04456
- MAE
- 0.01154
- MGE
- 0.01780
- MAE
- 0.00511
- MGE
- 0.01640✓
- MAE
- 0.00473✓
- MGE
- 0.02221✓
- MAE
- 0.00778✓
- MGE
- 0.03085
- MAE
- 0.01043
- MGE
- 0.04873✓
- MAE
- 0.01122✓
- MGE
- 0.09147
- MAE
- 0.05514
- MGE
- 0.01818✓
- MAE
- 0.00827✓
- MGE
- 0.02829
- MAE
- 0.01542
- MGE
- 0.01969✓
- MAE
- 0.01374✓
- MGE
- 0.02339
- MAE
- 0.01436
- MGE
- 0.05570✓
- MAE
- 0.02558✓
- MGE
- 0.07925
- MAE
- 0.12626
- MGE
- 0.01467✓
- MAE
- 0.00332✓
- MGE
- 0.02176
- MAE
- 0.00588
- MGE
- 0.02090✓
- MAE
- 0.00612✓
- MGE
- 0.03982
- MAE
- 0.01117
- MGE
- 0.00619✓
- MAE
- 0.00277✓
- MGE
- 0.01176
- MAE
- 0.00436
- MGE
- 0.01810✓
- MAE
- 0.00714✓
- MGE
- 0.03169
- MAE
- 0.01241
- MGE
- 0.03517✓
- MAE
- 0.00648✓
- MGE
- 0.04629
- MAE
- 0.00984
- MGE
- 0.02895
- MAE
- 0.02783✓
- MGE
- 0.02316✓
- MAE
- 0.02950
- MGE
- 0.01442✓
- MAE
- 0.00315✓
- MGE
- 0.02309
- MAE
- 0.00498
- MGE
- 0.01079✓
- MAE
- 0.00372✓
- MGE
- 0.01993
- MAE
- 0.00674
- MGE
- 0.01879✓
- MAE
- 0.00851✓
- MGE
- 0.02594
- MAE
- 0.01077
- MGE
- 0.04204✓
- MAE
- 0.07583✓
- MGE
- 0.06121
- MAE
- 0.25267
- MGE
- 0.09392
- MAE
- 0.06015
- MGE
- 0.05816✓
- MAE
- 0.04968✓
- MGE
- 0.07979
- MAE
- 0.09417
- MGE
- 0.06874✓
- MAE
- 0.03261✓
- MGE
- 0.09319✓
- MAE
- 0.01831✓
- MGE
- 0.13145
- MAE
- 0.02954
- MGE
- 0.03339✓
- MAE
- 0.01185✓
- MGE
- 0.03707
- MAE
- 0.02387
- MGE
- 0.04978✓
- MAE
- 0.01930✓
- MGE
- 0.05797
- MAE
- 0.02682
- MGE
- 0.01854✓
- MAE
- 0.00954✓
- MGE
- 0.02328
- MAE
- 0.01329
Methodology
- Alpha mattes are grayscale masks encoding per-pixel foreground probability. They capture fine edge detail (hair, fur, semi-transparent regions) that binary masks discard.
- Metrics are computed by comparing predicted mattes against a ground-truth reference: MGE measures edge sharpness and MAE measures pixel-level accuracy. Both are lower-is-better.
- Scored images come from Bench v0, where every image has a ground-truth matte. The benchmark page has the full methodology and per-image scores for every model (CSV / JSON).