Facial Landmark Detection on COFW-68
2.47NME Box (%)DTLD-s
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DTLD-s2022.10 | 2.47 | — | 65 | — | |
| DTLD-sPre-trained on 300W-LP-2D=false2022.08 | 2.47 | — | 65 | — | |
| SPIGAnumber of HG modules=42022.10 | 2.52 | — | 64.1 | 3.93 | |
| LUVLiPre-trained on 300W-LP-2D=true2022.08 | 2.57 | — | 63.4 | — | |
| GlomFace2022.10 | 2.69 | — | — | 4.22 | |
| KDNPre-trained on 300W-LP-2D=true2022.08 | 2.73 | — | 60.1 | — | |
| LUVLInumber of HG modules=82022.10 | 2.75 | — | 60.8 | 4.21 | |
| LUVLiPre-trained on 300W-LP-2D=false2022.08 | 2.75 | — | 60.8 | — | |
| SoftlabelPre-trained on 300W-LP-2D=true2022.08 | 2.92 | — | 57.9 | — | |
| 2D-FANPre-trained on 300W-LP-2D=true2022.08 | 2.95 | — | 57.5 | — | |
| SANPre-trained on 300W-LP-2D=true2022.08 | 3.5 | — | 51.9 | — | |
| AVS w/ SANBackbone=ITN-CPM2021.05 | — | 4.43 | — | — | |
| BarrelNetBackbone=ResNet-182021.05 | — | 4.41 | — | — | |
| BarrelNetBackbone=ResNet-1012021.05 | — | 4.27 | — | — | |
| DAGBackbone=HRNet-W182021.05 | — | 4.22 | — | — | |
| HG×1+SAAT2022.10 | — | — | — | 4.61 | |
| HRNetV2-W182022.10 | — | — | — | 5.06 | |
| LABBackbone=ResNet-182021.05 | — | 4.62 | — | — | |
| ODNBackbone=ResNet-182021.05 | — | 5.3 | — | — | |
| PIPNetBackbone=ResNet-182021.05 | — | 4.55 | — | — | |
| SDFL2022.10 | — | — | — | 4.18 | |
| SLD2022.10 | — | — | — | 4.22 | |
| SPLT2022.10 | — | — | — | 4.1 |