Full-body selfie generation on Collected selfie-to-full-body dataset 17 captures 1.0 (test)
0.187LPIPSTotal Selfie
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Total Selfievariant=Full Pipeline, target_pose=Ground Truth Canny Edge, evaluation_region=foreground only2023.08 | 0.187 | 0.708 | 17.01 | 139.4 | |
| Total Selfievariant=Ours-FU, face_undistortion=false, target_pose=Ground Truth Canny Edge, evaluation_region=foreground only2023.08 | 0.188 | 0.706 | 16.98 | 141.6 | |
| Total Selfievariant=Ours-AR, appearance_refinement=false, target_pose=Ground Truth Canny Edge, evaluation_region=foreground only2023.08 | 0.19 | 0.703 | 16.89 | 143.4 | |
| Total Selfievariant=Ours-FT-AR, per_capture_fine_tuning=false, appearance_refinement=false, target_pose=Ground Truth Canny Edge, evaluation_region=foreground only2023.08 | 0.218 | 0.691 | 14.19 | 154.1 | |
| DreamBoothtarget_pose=Ground Truth Canny Edge, evaluation_region=foreground only2023.08 | 0.224 | 0.663 | 13.61 | 159.2 | |
| Paint-By-Exampletarget_pose=Ground Truth Canny Edge, evaluation_region=foreground only2023.08 | 0.252 | 0.621 | 12.37 | 184.6 | |
| LaDI-VTONtarget_pose=Ground Truth Canny Edge, evaluation_region=foreground only2023.08 | 0.263 | 0.563 | 10.35 | 172.8 | |
| DisCotarget_pose=Ground Truth Canny Edge, evaluation_region=foreground only2023.08 | 0.287 | 0.586 | 12.13 | 245.7 |