- BiRefNet General
- Open weights · commercially permissiveLicense: MIT
- BiRefNet Portrait
- Open weights · commercially permissiveLicense: MIT
birefnet-general vs birefnet-portrait
Quantitative comparison: 34 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-portrait |
|---|---|---|
| MGE (Edge Quality) | 0.04117 ✓ | 0.05950 |
| MAE (Overall Accuracy) | 0.01632 ✓ | 0.05032 |
- MGE
- 0.03911✓
- MAE
- 0.00742✓
- MGE
- 0.14119
- MAE
- 0.13362
- MGE
- 0.05151✓
- MAE
- 0.02941✓
- MGE
- 0.06774
- MAE
- 0.22680
- MGE
- 0.03935✓
- MAE
- 0.00627✓
- MGE
- 0.04515
- MAE
- 0.00868
- MGE
- 0.01441✓
- MAE
- 0.00777
- MGE
- 0.01512
- MAE
- 0.00591✓
- MGE
- 0.02280✓
- MAE
- 0.00639✓
- MGE
- 0.06204
- MAE
- 0.03288
- MGE
- 0.05248
- MAE
- 0.01589✓
- MGE
- 0.05246✓
- MAE
- 0.02145
- MGE
- 0.03986✓
- MAE
- 0.01194✓
- MGE
- 0.04472
- MAE
- 0.01411
- MGE
- 0.06897✓
- MAE
- 0.01163✓
- MGE
- 0.16214
- MAE
- 0.10275
- MGE
- 0.02118✓
- MAE
- 0.00394
- MGE
- 0.02311
- MAE
- 0.00370✓
- MGE
- 0.03558
- MAE
- 0.01109
- MGE
- 0.02896✓
- MAE
- 0.00886✓
- MGE
- 0.04585✓
- MAE
- 0.00903✓
- MGE
- 0.08324
- MAE
- 0.04343
- MGE
- 0.04642✓
- MAE
- 0.00943✓
- MGE
- 0.05236
- MAE
- 0.01182
- MGE
- 0.02089
- MAE
- 0.00573
- MGE
- 0.01095✓
- MAE
- 0.00289✓
- MGE
- 0.03183✓
- MAE
- 0.00786✓
- MGE
- 0.06671
- MAE
- 0.02252
- MGE
- 0.05580✓
- MAE
- 0.01033✓
- MGE
- 0.15394
- MAE
- 0.04942
- MGE
- 0.01899✓
- MAE
- 0.00728✓
- MGE
- 0.01953
- MAE
- 0.00800
- MGE
- 0.01820✓
- MAE
- 0.01234✓
- MGE
- 0.01972
- MAE
- 0.01254
- MGE
- 0.08594✓
- MAE
- 0.02513✓
- MGE
- 0.13269
- MAE
- 0.22283
- MGE
- 0.01814✓
- MAE
- 0.00291✓
- MGE
- 0.03465
- MAE
- 0.00886
- MGE
- 0.02291✓
- MAE
- 0.00565✓
- MGE
- 0.07836
- MAE
- 0.32167
- MGE
- 0.00578✓
- MAE
- 0.00237
- MGE
- 0.00660
- MAE
- 0.00225✓
- MGE
- 0.01608✓
- MAE
- 0.00625
- MGE
- 0.01762
- MAE
- 0.00610✓
- MGE
- 0.03751✓
- MAE
- 0.00610✓
- MGE
- 0.09258
- MAE
- 0.02708
- MGE
- 0.03350
- MAE
- 0.02219
- MGE
- 0.02527✓
- MAE
- 0.01801✓
- MGE
- 0.01732
- MAE
- 0.00341
- MGE
- 0.01406✓
- MAE
- 0.00223✓
- MGE
- 0.00992
- MAE
- 0.00353
- MGE
- 0.00956✓
- MAE
- 0.00312✓
- MGE
- 0.01705✓
- MAE
- 0.00698✓
- MGE
- 0.01959
- MAE
- 0.00804
- MGE
- 0.05592
- MAE
- 0.07340
- MGE
- 0.03003✓
- MAE
- 0.00879✓
- MGE
- 0.11893
- MAE
- 0.06936
- MGE
- 0.04145✓
- MAE
- 0.01784✓
- MGE
- 0.13372✓
- MAE
- 0.09454✓
- MGE
- 0.15421
- MAE
- 0.21967
- MGE
- 0.09324✓
- MAE
- 0.01624✓
- MGE
- 0.22170
- MAE
- 0.10099
- MGE
- 0.03565
- MAE
- 0.01114
- MGE
- 0.03258✓
- MAE
- 0.00972✓
- MGE
- 0.05451
- MAE
- 0.01951
- MGE
- 0.04777✓
- MAE
- 0.01845✓
- MGE
- 0.02037
- MAE
- 0.01244
- MGE
- 0.01511✓
- MAE
- 0.00599✓
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).