- BiRefNet General
- Open weights · commercially permissiveLicense: MIT
- BiRefNet DIS
- Open weights · commercially permissiveLicense: MIT
birefnet-general vs birefnet-dis
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-dis |
|---|---|---|
| MGE (Edge Quality) | 0.04117 ✓ | 0.04864 |
| MAE (Overall Accuracy) | 0.01632 ✓ | 0.02170 |
- MGE
- 0.03911✓
- MAE
- 0.00742✓
- MGE
- 0.06176
- MAE
- 0.01648
- MGE
- 0.05151✓
- MAE
- 0.02941✓
- MGE
- 0.06272
- MAE
- 0.22221
- MGE
- 0.03935✓
- MAE
- 0.00627✓
- MGE
- 0.04235
- MAE
- 0.00635
- MGE
- 0.01441✓
- MAE
- 0.00777✓
- MGE
- 0.03041
- MAE
- 0.01116
- MGE
- 0.02280
- MAE
- 0.00639
- MGE
- 0.01743✓
- MAE
- 0.00260✓
- MGE
- 0.05248✓
- MAE
- 0.01589✓
- MGE
- 0.07135
- MAE
- 0.03480
- MGE
- 0.03986
- MAE
- 0.01194
- MGE
- 0.02727✓
- MAE
- 0.00605✓
- MGE
- 0.06897
- MAE
- 0.01163
- MGE
- 0.06233✓
- MAE
- 0.01052✓
- MGE
- 0.02118
- MAE
- 0.00394
- MGE
- 0.01658✓
- MAE
- 0.00351✓
- MGE
- 0.03558✓
- MAE
- 0.01109✓
- MGE
- 0.03814
- MAE
- 0.01310
- MGE
- 0.04585✓
- MAE
- 0.00903✓
- MGE
- 0.05923
- MAE
- 0.01355
- MGE
- 0.04642
- MAE
- 0.00943
- MGE
- 0.03749✓
- MAE
- 0.00757✓
- MGE
- 0.02089
- MAE
- 0.00573
- MGE
- 0.01993✓
- MAE
- 0.00569✓
- MGE
- 0.03183✓
- MAE
- 0.00786✓
- MGE
- 0.03281
- MAE
- 0.00791
- MGE
- 0.05580✓
- MAE
- 0.01033✓
- MGE
- 0.07007
- MAE
- 0.01449
- MGE
- 0.01899✓
- MAE
- 0.00728✓
- MGE
- 0.04690
- MAE
- 0.01818
- MGE
- 0.01820✓
- MAE
- 0.01234✓
- MGE
- 0.03823
- MAE
- 0.01632
- MGE
- 0.08594✓
- MAE
- 0.02513✓
- MGE
- 0.11482
- MAE
- 0.07979
- MGE
- 0.01814✓
- MAE
- 0.00291✓
- MGE
- 0.02413
- MAE
- 0.00931
- MGE
- 0.02291✓
- MAE
- 0.00565✓
- MGE
- 0.02987
- MAE
- 0.00839
- MGE
- 0.00578✓
- MAE
- 0.00237✓
- MGE
- 0.01324
- MAE
- 0.00409
- MGE
- 0.01608✓
- MAE
- 0.00625✓
- MGE
- 0.03872
- MAE
- 0.01174
- MGE
- 0.03751✓
- MAE
- 0.00610✓
- MGE
- 0.03971
- MAE
- 0.00622
- MGE
- 0.03350✓
- MAE
- 0.02219✓
- MGE
- 0.04507
- MAE
- 0.03097
- MGE
- 0.01732✓
- MAE
- 0.00341✓
- MGE
- 0.02127
- MAE
- 0.00391
- MGE
- 0.00992✓
- MAE
- 0.00353✓
- MGE
- 0.02381
- MAE
- 0.00589
- MGE
- 0.01705✓
- MAE
- 0.00698✓
- MGE
- 0.04428
- MAE
- 0.01370
- MGE
- 0.05592
- MAE
- 0.07340
- MGE
- 0.05441✓
- MAE
- 0.01532✓
- MGE
- 0.11893
- MAE
- 0.06936
- MGE
- 0.11042✓
- MAE
- 0.03693✓
- MGE
- 0.13372
- MAE
- 0.09454
- MGE
- 0.10653✓
- MAE
- 0.01926✓
- MGE
- 0.09324
- MAE
- 0.01624✓
- MGE
- 0.08656✓
- MAE
- 0.01763
- MGE
- 0.03565✓
- MAE
- 0.01114✓
- MGE
- 0.05509
- MAE
- 0.01970
- MGE
- 0.05451✓
- MAE
- 0.01951✓
- MGE
- 0.07268
- MAE
- 0.02857
- MGE
- 0.02037✓
- MAE
- 0.01244✓
- MGE
- 0.03834
- MAE
- 0.01603
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).