- BiRefNet General
- Open weights · commercially permissiveLicense: MIT
- IS-Net
- Open weights · commercially permissiveLicense: Apache-2.0
BiRefNet General vs IS-Net
Quantitative comparison: 60 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 | IS-Net |
|---|---|---|
| MGE (Edge Quality) | 0.04948 ✓ | 0.06117 |
| MAE (Overall Accuracy) | 0.04262 ✓ | 0.09403 |
- MGE
- 0.04048✓
- MAE
- 0.01865✓
- MGE
- 0.05443
- MAE
- 0.03368
- MGE
- 0.09594✓
- MAE
- 0.06799✓
- MGE
- 0.10000
- MAE
- 0.08035
Photo by Alexander Dummer on Unsplash
- MGE
- 0.02067✓
- MAE
- 0.00625✓
- MGE
- 0.05612
- MAE
- 0.06709
- MGE
- 0.02110
- MAE
- 0.00739✓
- MGE
- 0.01499✓
- MAE
- 0.00909
- MGE
- 0.17845
- MAE
- 0.20501✓
- MGE
- 0.17298✓
- MAE
- 0.50163
Photo by Andreas Rasmussen on Unsplash
- MGE
- 0.03450✓
- MAE
- 0.00754✓
- MGE
- 0.07715
- MAE
- 0.06843
- MGE
- 0.03391✓
- MAE
- 0.02906✓
- MGE
- 0.03648
- MAE
- 0.22345
- MGE
- 0.03369✓
- MAE
- 0.00657✓
- MGE
- 0.04986
- MAE
- 0.04518
- MGE
- 0.01534✓
- MAE
- 0.00932✓
- MGE
- 0.02351
- MAE
- 0.01318
- MGE
- 0.01776
- MAE
- 0.00662
- MGE
- 0.01766✓
- MAE
- 0.00398✓
- MGE
- 0.05518✓
- MAE
- 0.02445✓
- MGE
- 0.06260
- MAE
- 0.03047
- MGE
- 0.08626✓
- MAE
- 0.25936✓
- MGE
- 0.11043
- MAE
- 0.53085
- MGE
- 0.07192✓
- MAE
- 0.15422✓
- MGE
- 0.13812
- MAE
- 0.52014
Photo by Camila Quintero Franco on Unsplash
- MGE
- 0.02070✓
- MAE
- 0.00960✓
- MGE
- 0.02829
- MAE
- 0.02992
- MGE
- 0.06499✓
- MAE
- 0.01274✓
- MGE
- 0.06714
- MAE
- 0.01656
- MGE
- 0.01052✓
- MAE
- 0.00263✓
- MGE
- 0.01995
- MAE
- 0.00657
Photo by Carlo Sierra on Unsplash
- MGE
- 0.01420✓
- MAE
- 0.00333✓
- MGE
- 0.01933
- MAE
- 0.05531
- MGE
- 0.03582✓
- MAE
- 0.01162✓
- MGE
- 0.04392
- MAE
- 0.02047
- MGE
- 0.03407✓
- MAE
- 0.02479✓
- MGE
- 0.06349
- MAE
- 0.12138
Photo by Conor Samuel on Unsplash
- MGE
- 0.06431✓
- MAE
- 0.01097✓
- MGE
- 0.09092
- MAE
- 0.01702
- MGE
- 0.02967✓
- MAE
- 0.00716✓
- MGE
- 0.03653
- MAE
- 0.00970
- MGE
- 0.01780
- MAE
- 0.00511✓
- MGE
- 0.01503✓
- MAE
- 0.00604
- MGE
- 0.02221✓
- MAE
- 0.00778✓
- MGE
- 0.02520
- MAE
- 0.01193
- MGE
- 0.04873✓
- MAE
- 0.01122✓
- MGE
- 0.07112
- MAE
- 0.08266
- MGE
- 0.04918
- MAE
- 0.02346✓
- MGE
- 0.04507✓
- MAE
- 0.02659
- MGE
- 0.01367✓
- MAE
- 0.00401✓
- MGE
- 0.03035
- MAE
- 0.01258
Photo by Lars Bo Nielsen on Unsplash
- MGE
- 0.07271✓
- MAE
- 0.09303✓
- MGE
- 0.07748
- MAE
- 0.09633
Photo by Mayer Tawfik on Unsplash
- MGE
- 0.01818✓
- MAE
- 0.00827✓
- MGE
- 0.04351
- MAE
- 0.02656
- MGE
- 0.01969✓
- MAE
- 0.01374✓
- MGE
- 0.03194
- MAE
- 0.02655
- MGE
- 0.03639✓
- MAE
- 0.01379✓
- MGE
- 0.04549
- MAE
- 0.02499
Photo by Niko Tsviliov on Unsplash
- MGE
- 0.23620
- MAE
- 0.14015
- MGE
- 0.19124✓
- MAE
- 0.13921✓
Photo by Norbert Braun on Unsplash
- MGE
- 0.01254✓
- MAE
- 0.00193✓
- MGE
- 0.04270
- MAE
- 0.06365
- MGE
- 0.05570✓
- MAE
- 0.02558✓
- MGE
- 0.10276
- MAE
- 0.25500
- MGE
- 0.06054
- MAE
- 0.09974✓
- MGE
- 0.05982✓
- MAE
- 0.10379
- MGE
- 0.03188✓
- MAE
- 0.02975✓
- MGE
- 0.13998
- MAE
- 0.28903
- MGE
- 0.01467✓
- MAE
- 0.00332✓
- MGE
- 0.04162
- MAE
- 0.05030
- MGE
- 0.34926
- MAE
- 0.46453
- MGE
- 0.32406✓
- MAE
- 0.46089✓
- MGE
- 0.02090✓
- MAE
- 0.00612✓
- MGE
- 0.03081
- MAE
- 0.01417
- MGE
- 0.00619✓
- MAE
- 0.00277✓
- MGE
- 0.01433
- MAE
- 0.01354
- MGE
- 0.01810✓
- MAE
- 0.00714✓
- MGE
- 0.03714
- MAE
- 0.02013
- MGE
- 0.03517✓
- MAE
- 0.00648✓
- MGE
- 0.03730
- MAE
- 0.00761
- MGE
- 0.02895
- MAE
- 0.02783✓
- MGE
- 0.02320✓
- MAE
- 0.03234
- MGE
- 0.01442✓
- MAE
- 0.00315✓
- MGE
- 0.01773
- MAE
- 0.00400
- MGE
- 0.01079✓
- MAE
- 0.00372✓
- MGE
- 0.01768
- MAE
- 0.00610
- MGE
- 0.01879✓
- MAE
- 0.00851✓
- MGE
- 0.03757
- MAE
- 0.02450
- MGE
- 0.00747✓
- MAE
- 0.00115✓
- MGE
- 0.00916
- MAE
- 0.00133
- MGE
- 0.04204✓
- MAE
- 0.07583
- MGE
- 0.04486
- MAE
- 0.02457✓
- MGE
- 0.09392✓
- MAE
- 0.06015
- MGE
- 0.09884
- MAE
- 0.05848✓
- MGE
- 0.17273
- MAE
- 0.03358
- MGE
- 0.11415✓
- MAE
- 0.02523✓
- MGE
- 0.07979
- MAE
- 0.09417
- MGE
- 0.07771✓
- MAE
- 0.02836✓
- MGE
- 0.09319✓
- MAE
- 0.01831✓
- MGE
- 0.09445
- MAE
- 0.01978
- MGE
- 0.01951
- MAE
- 0.00290
- MGE
- 0.01673✓
- MAE
- 0.00255✓
- MGE
- 0.03390✓
- MAE
- 0.00917✓
- MGE
- 0.08999
- MAE
- 0.04431
- MGE
- 0.01230✓
- MAE
- 0.01388✓
- MGE
- 0.02023
- MAE
- 0.05742
Photo by Vinicius Wiesehofer on Unsplash
- MGE
- 0.04915✓
- MAE
- 0.06760✓
- MGE
- 0.06374
- MAE
- 0.70768
- MGE
- 0.04032✓
- MAE
- 0.21379✓
- MGE
- 0.04792
- MAE
- 0.32880
- MGE
- 0.03049
- MAE
- 0.02974
- MGE
- 0.02564✓
- MAE
- 0.01765✓
- MGE
- 0.03339✓
- MAE
- 0.01185✓
- MGE
- 0.06352
- MAE
- 0.02465
- MGE
- 0.04978✓
- MAE
- 0.01930✓
- MGE
- 0.08990
- MAE
- 0.07705
- MGE
- 0.01854✓
- MAE
- 0.00954✓
- MGE
- 0.02635
- MAE
- 0.02086
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).
Photo Credits
Source images are courtesy of the photographers below via Unsplash.