Research map / Image generation / Editing and style transfer
PIRenderer: Controllable Portrait Image Generation via Semantic Neural Rendering
Yurui Ren, Ge Li, Yuanqi Chen, Thomas H. Li, Shan Liu
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 |
|---|---|---|---|
| use a window of continuous frames of 3DMM coefficients instead of a single frame as motion descriptor (ablation) | PIRenderer · Same-identity reconstruction | FID: 8.752 -> 8.260 | Improved |
| use a window of continuous frames of 3DMM coefficients instead of a single frame as motion descriptor (ablation) | PIRenderer · Same-identity reconstruction | AED: 0.1169 -> 0.1106 | Improved |
| use a window of continuous frames of 3DMM coefficients instead of a single frame as motion descriptor (ablation) | PIRenderer · Same-identity reconstruction | LPIPS: 0.1370 -> 0.1285 | Improved |
| use a window of continuous frames of 3DMM coefficients instead of a single frame as motion descriptor (ablation) | PIRenderer · Same-identity reconstruction | APD: 0.0182 -> 0.0164 | Improved |
| use fully disentangled 3DMM motion descriptor instead of keypoints (FOMM) | PIRenderer · Cross-identity motion imitation | FID: 38.92 -> 28.18 | Improved |
| use fully disentangled 3DMM motion descriptor instead of landmarks (GFLA) | PIRenderer · Cross-identity motion imitation | AED: 0.2599 -> 0.2304 | Improved |
| use PIRenderer instead of FOMM | PIRenderer · Same-identity reconstruction | LPIPS: 0.1220 -> 0.1285 | Hurt |
| use PIRenderer instead of StyleRig (StyleGAN generator) | PIRenderer · Portrait editing (StyleGAN images) | FID: 47.37 -> 65.97 | Hurt |
| use PIRenderer instead of StyleRig | PIRenderer · Portrait editing (StyleGAN images) | AED: 0.316 -> 0.257 | Improved |
| use PIRenderer instead of StyleRig | PIRenderer · Portrait editing (StyleGAN images) | APD: 0.0919 -> 0.0252 | Improved |
| use PIRenderer instead of FOMM | PIRenderer · Cross-identity motion imitation | JND fooling rate: 19.4% -> 25.5% | Improved |
Later papers that compared with this method
1 later accepted paper reports PIRenderer as a baseline in their results tables; 0 of them report a better number on most of the results they share with it.
- Face Adapter for Pre-Trained Diffusion Models with Fine-Grained ID and Attribute Control (ECCV 2024 · better on 1 of 2 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- RealPortrait: Realistic Portrait Animation with Diffusion Transformers (AAAI 2025 · similarity 0.86)
- Hallo3: Highly Dynamic and Realistic Portrait Image Animation with Video Diffusion Transformer (CVPR 2025 · similarity 0.85)
- GoHD: Gaze-oriented and Highly Disentangled Portrait Animation with Rhythmic Poses and Realistic Expressions (AAAI 2025 · similarity 0.84)
- HunyuanPortrait: Implicit Condition Control for Enhanced Portrait Animation (CVPR 2025 · similarity 0.84)
- InstructPix2NeRF: Instructed 3D Portrait Editing from a Single Image (ICLR 2024 · similarity 0.84)
- ControlFace: Harnessing Facial Parametric Control for Face Rigging (CVPR 2025 · similarity 0.84)
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.
