Research map / Image generation / Editing and style transfer
MarioNETte: Few-Shot Face Reenactment Preserving Identity of Unseen Targets
Sungjoo Ha, Martin Kersner, Beomsu Kim, Seokjun Seo, Dongyoung Kim
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add the landmark transformer (ablation) | MarioNETte · CelebV different-identity reenactment, 1 target | CSIM: 0.520 -> 0.568 | Improved |
| Add the landmark transformer (ablation) | MarioNETte · CelebV different-identity reenactment, 8 targets | CSIM: 0.608 -> 0.661 | Improved |
| Add the landmark transformer (ablation) | MarioNETte · CelebV different-identity reenactment | PRMSE: 3.41 -> 3.70; 3.26 -> 3.57 | Hurt |
| Replace AdaIN with image attention (ablation) | MarioNETte · CelebV different-identity reenactment, 1 target | CSIM: 0.063 -> 0.333 | Improved |
| Replace AdaIN with image attention (ablation) | MarioNETte · CelebV different-identity reenactment, 8 targets | CSIM: 0.069 -> 0.472 | Improved |
| Add target feature alignment instead of using attention alone (ablation) | MarioNETte · CelebV different-identity reenactment, 1 target | CSIM: 0.333 -> 0.530 | Improved |
| Add target feature alignment instead of using attention alone (ablation) | MarioNETte · CelebV different-identity reenactment, 8 targets | CSIM: 0.472 -> 0.605 | Improved |
| Use alignment without image attention (ablation) | MarioNETte · CelebV different-identity reenactment | PRMSE: 3.44; 3.27 | Hurt |
| Combine image attention and target feature alignment (ablation) | MarioNETte · CelebV different-identity reenactment | CSIM: 0.530 -> 0.520; 0.605 -> 0.608 | Mixed |
| Add image attention instead of AdaIN (ablation) | MarioNETte · VoxCeleb1 self-reenactment, 1 target | CSIM: 0.183 -> 0.611 | Improved |
| Use target feature alignment only instead of image attention only (ablation) | MarioNETte · VoxCeleb1 self-reenactment, 8 targets | PRMSE: 2.909 -> 2.563 | Improved |
| Add image attention and target feature alignment (ablation) | MarioNETte · CelebV different-identity reenactment | identity preservation (CSIM) | Improved |
Closest papers by meaning
- FReeNet: Multi-Identity Face Reenactment (CVPR 2020 · similarity 0.87)
- Designing One Unified Framework for High-Fidelity Face Reenactment and Swapping (ECCV 2022 · similarity 0.85)
- FaceMe: Robust Blind Face Restoration with Personal Identification (AAAI 2025 · similarity 0.84)
- HiFi-Portrait: Zero-shot Identity-preserved Portrait Generation with High-fidelity Multi-face Fusion (CVPR 2025 · similarity 0.84)
- VOODOO 3D: Volumetric Portrait Disentanglement For One-Shot 3D Head Reenactment (CVPR 2024 · similarity 0.83)
- FaithfulFaces: Pose-Faithful Facial Identity Preservation for Text-to-Video Generation (NeurIPS 2026 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
