- U²-Net
- Open weights · commercially permissiveLicense: Apache-2.0
- U²-NetP
- Open weights · commercially permissiveLicense: Apache-2.0
u2net vs u2netp
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 | u2net | u2netp |
|---|---|---|
| MGE (Edge Quality) | 0.05404 ✓ | 0.05481 |
| MAE (Overall Accuracy) | 0.06673 ✓ | 0.07206 |



- MGE
- 0.09509✓
- MAE
- 0.08100✓
- MGE
- 0.09894
- MAE
- 0.10206



- MGE
- 0.03862✓
- MAE
- 0.22676
- MGE
- 0.03897
- MAE
- 0.22444✓



- MGE
- 0.06007
- MAE
- 0.03098✓
- MGE
- 0.05958✓
- MAE
- 0.06814



- MGE
- 0.03274
- MAE
- 0.01732✓
- MGE
- 0.02975✓
- MAE
- 0.01762



- MGE
- 0.04736✓
- MAE
- 0.01496✓
- MGE
- 0.05210
- MAE
- 0.05059



- MGE
- 0.06842
- MAE
- 0.04546
- MGE
- 0.06157✓
- MAE
- 0.03072✓



- MGE
- 0.03513
- MAE
- 0.03698
- MGE
- 0.02327✓
- MAE
- 0.01315✓



- MGE
- 0.11366✓
- MAE
- 0.07027
- MGE
- 0.11602
- MAE
- 0.06524✓



- MGE
- 0.04141✓
- MAE
- 0.14397✓
- MGE
- 0.04206
- MAE
- 0.23785



- MGE
- 0.04762✓
- MAE
- 0.02104✓
- MGE
- 0.05232
- MAE
- 0.02940



- MGE
- 0.06526✓
- MAE
- 0.02066✓
- MGE
- 0.06844
- MAE
- 0.02526



- MGE
- 0.02158✓
- MAE
- 0.00659✓
- MGE
- 0.02605
- MAE
- 0.00943



- MGE
- 0.03629✓
- MAE
- 0.03255✓
- MGE
- 0.04282
- MAE
- 0.04616



- MGE
- 0.09893
- MAE
- 0.09053
- MGE
- 0.09621✓
- MAE
- 0.08122✓



- MGE
- 0.03523✓
- MAE
- 0.02607✓
- MGE
- 0.03691
- MAE
- 0.03427



- MGE
- 0.02820✓
- MAE
- 0.02370
- MGE
- 0.02867
- MAE
- 0.02323✓



- MGE
- 0.07998✓
- MAE
- 0.29806✓
- MGE
- 0.08134
- MAE
- 0.31521



- MGE
- 0.04300
- MAE
- 0.07894✓
- MGE
- 0.04015✓
- MAE
- 0.08022



- MGE
- 0.06347✓
- MAE
- 0.03625✓
- MGE
- 0.06854
- MAE
- 0.05066



- MGE
- 0.02008✓
- MAE
- 0.01485✓
- MGE
- 0.02268
- MAE
- 0.01677



- MGE
- 0.03800✓
- MAE
- 0.01532✓
- MGE
- 0.03923
- MAE
- 0.02272



- MGE
- 0.06015✓
- MAE
- 0.14384
- MGE
- 0.06267
- MAE
- 0.12973✓



- MGE
- 0.03059✓
- MAE
- 0.04455✓
- MGE
- 0.03479
- MAE
- 0.07184



- MGE
- 0.03554✓
- MAE
- 0.00845✓
- MGE
- 0.03858
- MAE
- 0.01002



- MGE
- 0.02402✓
- MAE
- 0.00790✓
- MGE
- 0.02480
- MAE
- 0.00904



- MGE
- 0.03196✓
- MAE
- 0.02429✓
- MGE
- 0.03431
- MAE
- 0.03237



- MGE
- 0.04327
- MAE
- 0.02803
- MGE
- 0.03782✓
- MAE
- 0.01914✓



- MGE
- 0.06458
- MAE
- 0.10452
- MGE
- 0.05506✓
- MAE
- 0.07337✓



- MGE
- 0.07450✓
- MAE
- 0.05224✓
- MGE
- 0.07959
- MAE
- 0.06058



- MGE
- 0.16225✓
- MAE
- 0.11610
- MGE
- 0.16708
- MAE
- 0.10251✓



- MGE
- 0.04612
- MAE
- 0.07316
- MGE
- 0.04383✓
- MAE
- 0.04392✓



- MGE
- 0.07274✓
- MAE
- 0.23800✓
- MGE
- 0.07506
- MAE
- 0.24634



- MGE
- 0.02746✓
- MAE
- 0.02864✓
- MGE
- 0.02951
- MAE
- 0.03488
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).