3D face reconstruction on benchmark High-Quality (HQ) 1.0
1.58Median Error (mm)RingNet
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| RingNet3D supervision=false, Evaluation subset=1427 images2019.05 | 1.58 | 2.02 | 1.69 | |
| RingNetBackbone=ResNet-502020.08 | 1.58 | 2.02 | 1.69 | |
| PRNet3D supervision=true, Evaluation subset=1427 images2019.05 | 1.6 | 2.06 | 1.79 | |
| 3DMM-CNN3D supervision=true, Evaluation subset=1427 images2019.05 | 1.85 | 2.29 | 1.88 | |
| ExPose face sub-network fBackbone=ResNet-182020.08 | 1.91 | 2.42 | 2.03 | |
| Extreme3D3D supervision=true, Evaluation subset=1427 images2019.05 | 2.37 | 3.58 | 6.75 |