BiRefNet General vs IS-Net
37 examples, 37 with ground truth. 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) | 0.06476 ✓ | 0.08337 |
| MAE (Overall Accuracy) | 0.05694 ✓ | 0.08659 |
| Connectivity | 0.05636 ✓ | 0.08562 |
- MGE
- 0.05164✓
- MAE
- 0.01051✓
- Conn.
- 0.00860✓
- MGE
- 0.12449
- MAE
- 0.07297
- Conn.
- 0.06980
- MGE
- 0.06089✓
- MAE
- 0.03207✓
- Conn.
- 0.03135✓
- MGE
- 0.06655
- MAE
- 0.22381
- Conn.
- 0.22360
- MGE
- 0.06130✓
- MAE
- 0.01128✓
- Conn.
- 0.01096✓
- MGE
- 0.07657
- MAE
- 0.06137
- Conn.
- 0.06394
- MGE
- 0.01673✓
- MAE
- 0.00895✓
- Conn.
- 0.00713✓
- MGE
- 0.02615
- MAE
- 0.01341
- Conn.
- 0.01185
- MGE
- 0.02841✓
- MAE
- 0.00747
- Conn.
- 0.00708
- MGE
- 0.03013
- MAE
- 0.00528✓
- Conn.
- 0.00464✓
- MGE
- 0.04911✓
- MAE
- 0.01525✓
- Conn.
- 0.01362✓
- MGE
- 0.06105
- MAE
- 0.02702
- Conn.
- 0.02405
- MGE
- 0.03878
- MAE
- 0.01145✓
- Conn.
- 0.01170✓
- MGE
- 0.03522✓
- MAE
- 0.02263
- Conn.
- 0.02122
- MGE
- 0.08066✓
- MAE
- 0.01395✓
- Conn.
- 0.01311✓
- MGE
- 0.08820
- MAE
- 0.01739
- Conn.
- 0.01604
- MGE
- 0.02619✓
- MAE
- 0.00475✓
- Conn.
- 0.00359✓
- MGE
- 0.04268
- MAE
- 0.06100
- Conn.
- 0.05949
- MGE
- 0.03370✓
- MAE
- 0.01148✓
- Conn.
- 0.01015✓
- MGE
- 0.05411
- MAE
- 0.02930
- Conn.
- 0.02823
- MGE
- 0.07350
- MAE
- 0.01472✓
- Conn.
- 0.01452✓
- MGE
- 0.06718✓
- MAE
- 0.01971
- Conn.
- 0.01778
- MGE
- 0.04902✓
- MAE
- 0.00965✓
- Conn.
- 0.00845✓
- MGE
- 0.05507
- MAE
- 0.01320
- Conn.
- 0.01178
- MGE
- 0.01263✓
- MAE
- 0.00370✓
- Conn.
- 0.00266✓
- MGE
- 0.01957
- MAE
- 0.00658
- Conn.
- 0.00563
- MGE
- 0.04433✓
- MAE
- 0.00942✓
- Conn.
- 0.00934✓
- MGE
- 0.04879
- MAE
- 0.01407
- Conn.
- 0.01423
- MGE
- 0.07464✓
- MAE
- 0.01590✓
- Conn.
- 0.01520✓
- MGE
- 0.12660
- MAE
- 0.08300
- Conn.
- 0.08018
- MGE
- 0.01940✓
- MAE
- 0.00869✓
- Conn.
- 0.00767✓
- MGE
- 0.03633
- MAE
- 0.02418
- Conn.
- 0.02425
- MGE
- 0.01867✓
- MAE
- 0.01305✓
- Conn.
- 0.01152✓
- MGE
- 0.03501
- MAE
- 0.02718
- Conn.
- 0.02617
- MGE
- 0.08035✓
- MAE
- 0.02283✓
- Conn.
- 0.02099✓
- MGE
- 0.17796
- MAE
- 0.27290
- Conn.
- 0.27664
- MGE
- 0.02727✓
- MAE
- 0.00459✓
- Conn.
- 0.00401✓
- MGE
- 0.08141
- MAE
- 0.05662
- Conn.
- 0.05412
- MGE
- 0.34795
- MAE
- 0.46062✓
- Conn.
- 0.47012✓
- MGE
- 0.33448✓
- MAE
- 0.46974
- Conn.
- 0.48246
- MGE
- 0.03023✓
- MAE
- 0.00675✓
- Conn.
- 0.00548✓
- MGE
- 0.05076
- MAE
- 0.01701
- Conn.
- 0.01449
- MGE
- 0.01159✓
- MAE
- 0.00830✓
- Conn.
- 0.00688✓
- MGE
- 0.01823
- MAE
- 0.01416
- Conn.
- 0.01331
- MGE
- 0.01939✓
- MAE
- 0.00754✓
- Conn.
- 0.00501✓
- MGE
- 0.03478
- MAE
- 0.02152
- Conn.
- 0.01752
- MGE
- 0.05645✓
- MAE
- 0.00930✓
- Conn.
- 0.00850✓
- MGE
- 0.06092
- MAE
- 0.01037
- Conn.
- 0.00923
- MGE
- 0.03070✓
- MAE
- 0.01848✓
- Conn.
- 0.01790✓
- MGE
- 0.03269
- MAE
- 0.03657
- Conn.
- 0.03447
- MGE
- 0.01599✓
- MAE
- 0.00337✓
- Conn.
- 0.00219✓
- MGE
- 0.02644
- MAE
- 0.00537
- Conn.
- 0.00451
- MGE
- 0.01045✓
- MAE
- 0.00402✓
- Conn.
- 0.00262✓
- MGE
- 0.01871
- MAE
- 0.00612
- Conn.
- 0.00494
- MGE
- 0.02125✓
- MAE
- 0.00843✓
- Conn.
- 0.00666✓
- MGE
- 0.03361
- MAE
- 0.02629
- Conn.
- 0.02443
- MGE
- 0.05709✓
- MAE
- 0.02184✓
- Conn.
- 0.02052✓
- MGE
- 0.06968
- MAE
- 0.02905
- Conn.
- 0.02494
- MGE
- 0.08856
- MAE
- 0.06629
- Conn.
- 0.06887
- MGE
- 0.08788✓
- MAE
- 0.05431✓
- Conn.
- 0.05123✓
- MGE
- 0.45101✓
- MAE
- 0.90677
- Conn.
- 0.91173
- MGE
- 0.56277
- MAE
- 0.90155✓
- Conn.
- 0.90653✓
- MGE
- 0.12912
- MAE
- 0.08314
- Conn.
- 0.08233
- MGE
- 0.10144✓
- MAE
- 0.02800✓
- Conn.
- 0.02763✓
- MGE
- 0.12194
- MAE
- 0.02224✓
- Conn.
- 0.02147
- MGE
- 0.11987✓
- MAE
- 0.02251
- Conn.
- 0.02072✓
- MGE
- 0.06271✓
- MAE
- 0.21489✓
- Conn.
- 0.21447✓
- MGE
- 0.08196
- MAE
- 0.35117
- Conn.
- 0.35466
- MGE
- 0.03058✓
- MAE
- 0.00983✓
- Conn.
- 0.00802✓
- MGE
- 0.06492
- MAE
- 0.02734
- Conn.
- 0.02722
- MGE
- 0.04706✓
- MAE
- 0.01840✓
- Conn.
- 0.01632✓
- MGE
- 0.10094
- MAE
- 0.10320
- Conn.
- 0.09046
- MGE
- 0.01689✓
- MAE
- 0.00690✓
- Conn.
- 0.00444✓
- MGE
- 0.03163
- MAE
- 0.02776
- Conn.
- 0.02542
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, MAE measures pixel-level accuracy, and Connectivity penalises fragmented foreground regions. All three are lower-is-better.
- Scores are shown as ? when no ground-truth matte is available for that image. See the alpha matting evaluation benchmark for details on the evaluation methodology.