Research map / 3D vision / Face and shape
HumanLiff: Layer-wise 3D Human Diffusion Model
Shoukang Hu, Fangzhou Hong, Tao Hu, Liang Pan, Haiyi Mei, Weiye Xiao, Lei Yang, Ziwei 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 |
|---|---|---|---|
| add 3D condition UNet encoder instead of AdaGN (ablation) | HumanLiff · SynBody layer-wise (Layer 1) | FID: 28.37 -> 25.09 | Improved |
| add 3D condition UNet encoder instead of AdaGN (ablation) | HumanLiff · SynBody layer-wise (Layer 1) | L-PSNR: 20.96 -> 28.14 | Improved |
| add 3D condition UNet encoder instead of AdaGN (ablation) | HumanLiff · SynBody layer-wise (Layer 3) | L-PSNR: 22.93 -> 29.87 | Improved |
| add 3D condition UNet encoder instead of cross-attention (ablation) | HumanLiff · SynBody layer-wise (Layer 0) | FID: 26.62 -> 22.28 | Improved |
| add 3D condition UNet encoder instead of cross-attention (ablation) | HumanLiff · SynBody layer-wise (Layer 1) | L-PSNR: 27.18 -> 28.14 | Improved |
| add 3D condition UNet encoder instead of cross-attention (ablation) | HumanLiff · SynBody layer-wise (Layer 3) | FID: 27.94 -> 25.20 | Improved |
| add tri-plane shift (ablation) | layer-wise tri-plane fitting · SynBody tri-plane fitting | PSNR: 31.35 -> 31.97 | Improved |
| add tri-plane shift (ablation) | layer-wise tri-plane fitting · SynBody tri-plane fitting | LPIPS: 0.05 -> 0.04 | Improved |
| add tri-plane shift (ablation) | layer-wise tri-plane fitting · SynBody tri-plane fitting | SSIM: 0.97 -> 0.98 | Improved |
| add AdaGN conditioning on 1D latent (Rodin) in place of 3D UNet encoder (ablation) | Rodin vs HumanLiff · SynBody layer-wise (Layer 1) | L-PSNR: 28.14 -> 19.54 | Hurt |
Later papers that compared with this method
1 later accepted paper reports HumanLiff as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Disentangled Clothed Avatar Generation with Layered Representation (ICCV 2025 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- EVA3D — EVA3D: Compositional 3D Human Generation from 2D Image Collections (ICLR 2023) · better on 4 of 4 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
- HumanRef: Single Image to 3D Human Generation via Reference-Guided Diffusion (CVPR 2024 · similarity 0.90)
- HumanNorm: Learning Normal Diffusion Model for High-quality and Realistic 3D Human Generation (CVPR 2024 · similarity 0.89)
- DreamFit: Garment-Centric Human Generation via a Lightweight Anything-Dressing Encoder (AAAI 2025 · similarity 0.89)
- R-Cyclic Diffuser: Reductive and Cyclic Latent Diffusion for 3D Clothed Human Digitalization (CVPR 2024 · similarity 0.88)
- FreeCloth: Free-form Generation Enhances Challenging Clothed Human Modeling (CVPR 2025 · similarity 0.88)
- DeClotH: Decomposable 3D Cloth and Human Body Reconstruction from a Single Image (CVPR 2025 · similarity 0.88)
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.
