Research map / 3D vision / Human pose estimation
Unselfie: Translating Selfies to Neutral-pose Portraits in the Wild
Liqian Ma, Zhe Lin, Connelly Barnes, Alexei A Efros, Jingwan Lu
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 perceptual loss L_P to the composition network (ablation) | composition network · Unselfie test set (500 selfies) | FID: 82.09 -> 71.93 | Improved |
| add perceptual loss L_P to the composition network (ablation) | composition network · Unselfie test set (500 selfies) | KID: 0.019 -> 0.014 | Improved |
| add deformable skip-connections (ablation) | composition network · Unselfie test set (500 selfies) | FID: 73.87 -> 71.93 | Improved |
| add deformable skip-connections (ablation) | composition network · Unselfie test set (500 selfies) | KID: 0.017 -> 0.014 | Improved |
| use gated convolution instead of normal convolution (ablation) | composition network · Unselfie test set (500 selfies) | FID: 72.89 -> 71.93 | Improved |
| use gated convolution instead of normal convolution (ablation) | composition network · Unselfie test set (500 selfies) | KID: 0.014 -> 0.014 | No clear effect |
| use the two-step pose search (body index then UV distance) instead of either distance alone (ablation) | nearest pose search · qualitative | retrieved pose quality | Improved |
| use the full pipeline instead of DPIG | unselfie pipeline · Unselfie test set (500 selfies) | FID: 88.27 -> 71.93 | Improved |
| use the full pipeline instead of PATN | unselfie pipeline · Unselfie test set (500 selfies) | FID: 104.74 -> 71.93 | Improved |
| pre-fill background holes with an off-the-shelf inpainting network | unselfie pipeline · qualitative | visual quality | Mixed |
Closest papers by meaning
- Pro-Pose: Unpaired Full-Body Portrait Synthesis via Canonical UV Maps (ECCV 2026 · similarity 0.87)
- IMAGPose: A Unified Conditional Framework for Pose-Guided Person Generation (NeurIPS 2024 · similarity 0.83)
- MagiCapture: High-Resolution Multi-Concept Portrait Customization (AAAI 2024 · similarity 0.83)
- UniHuman: A Unified Model For Editing Human Images in the Wild (CVPR 2024 · similarity 0.83)
- FRESA: Feedforward Reconstruction of Personalized Skinned Avatars from Few Images (CVPR 2025 · similarity 0.82)
- MagicPose: Realistic Human Poses and Facial Expressions Retargeting with Identity-aware Diffusion (ICML 2024 · similarity 0.82)
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.
