Face Parsing on LaPa
92.32Mean F1 ScoreFaRL
Evaluation Results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| FaRLPre-training Dataset=LAION-FACE, Pre-training Scale=20M, Supervision Source=face images + web text, Evaluation Protocol=Frozen backbone + head fine-tuning2021.12 | 92.32 | — | — | — | — | — | — | — | — | — | — | — | |
| CLIPPre-training Dataset=WIT, Pre-training Scale=400M, Supervision Source=images + web text, Evaluation Protocol=Frozen backbone + head fine-tuning2021.12 | 92.21 | — | — | — | — | — | — | — | — | — | — | — | |
| DeiTPre-training Dataset=ImageNet-1K, Pre-training Scale=1.3M, Supervision Source=images + human labels, Evaluation Protocol=Frozen backbone + head fine-tuning2021.12 | 92 | — | — | — | — | — | — | — | — | — | — | — | |
| MoCo v3Pre-training Dataset=ImageNet-1K, Pre-training Scale=1.3M, Supervision Source=images, Evaluation Protocol=Frozen backbone + head fine-tuning2021.12 | 91.86 | — | — | — | — | — | — | — | — | — | — | — | |
| ViTPre-training Dataset=ImageNet-22K, Pre-training Scale=14M, Supervision Source=images + human labels, Evaluation Protocol=Frozen backbone + head fine-tuning2021.12 | 91.61 | — | — | — | — | — | — | — | — | — | — | — | |
| BEiTPre-training Dataset=ImageNet-22K, Pre-training Scale=14M, Supervision Source=images, Evaluation Protocol=Frozen backbone + head fine-tuning2021.12 | 91.29 | — | — | — | — | — | — | — | — | — | — | — | |
| Face TransformerPre-training Dataset=MS1MV3, Pre-training Scale=5.1M, Supervision Source=face images + human labels, Evaluation Protocol=Frozen backbone + head fine-tuning2021.12 | 91.09 | — | — | — | — | — | — | — | — | — | — | — | |
| Liu et al.2021.02 | — | 88.1 | 88 | 84.4 | 87.6 | 85.7 | 95.5 | 87.7 | 87.6 | 97.2 | 96.3 | 89.8 | |
| RTNet2021.02 | — | 91.5 | 90.9 | 88.7 | 90.5 | 90.5 | 96.9 | 90.1 | 89.1 | 97.8 | 96.5 | 92.5 | |
| Te et al.2021.02 | — | 89.5 | 90 | 88.1 | 90 | 89 | 97.1 | 86.5 | 87 | 97.3 | 96.2 | 91.1 | |
| Zhao et al.2021.02 | — | 86.3 | 86 | 83.6 | 86.9 | 84.7 | 94.8 | 86.8 | 86.9 | 93.5 | 94.1 | 88.4 |