Research map / Image generation / Editing and style transfer
UnitedHuman: Harnessing Multi-Source Data for High-Resolution Human Generation
Jianglin Fu, Shikai Li, Yuming Jiang, Kwan-Yee Lin, Wayne Wu, 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 SHHQ-SR and DeepFashion partial datasets to 10K SHHQ-HR (ablation) | UnitedHuman · SHHQ 2048px | mean kFID: 25.48 -> 21.98 | Improved |
| add CelebA face dataset (ablation) | UnitedHuman · SHHQ 2048px | mean kFID: 21.98 -> 21.50 | Mixed |
| add synthetic DART hand dataset (ablation) | UnitedHuman · SHHQ 2048px | mean kFID: 21.50 -> 19.56 | Improved |
| align partial images with SMPL-based Multi-source Spatial Transformer instead of 2D keypoints (ablation) | UnitedHuman · SHHQ 2048px | mean kFID: 22.45 -> 19.56 | Improved |
| align with SMPL instead of pose-mapping MLP (ablation) | UnitedHuman · SHHQ 2048px | mean kFID: 27.81 -> 19.56 | Improved |
| add pixel loss from low-resolution stage-1 teacher for global structure (ablation) | UnitedHuman · SHHQ 2048px | mean kFID: reduction of 20.7 points in mean kFID | Improved |
| add pixel-wise discriminator with CutMix consistency (ablation) | UnitedHuman · SHHQ 2048px | mean kFID: kFID and pFID dropped to 19.56 and 18.94 | Improved |
| UnitedHuman with 10K full-body HR images vs StyleGAN-Human with 100K | UnitedHuman · SHHQ 2048px | mean kFID: 20.25 -> 19.56 | Improved |
| multi-source training vs StyleGAN-Human holistic training | UnitedHuman · SHHQ full-body | FID: 13.81 -> 15.37 | Hurt |
| multi-source training vs StyleGAN-Human | UnitedHuman · SHHQ 2048px | Recall: 0.65 -> 0.61 | Hurt |
Closest papers by meaning
- HyperHuman: Hyper-Realistic Human Generation with Latent Structural Diffusion (ICLR 2024 · similarity 0.89)
- CapHuman: Capture Your Moments in Parallel Universes (CVPR 2024 · similarity 0.87)
- RealisHuman: A Two-Stage Approach for Refining Malformed Human Parts in Generated Images (AAAI 2025 · similarity 0.87)
- MultiHuman-Testbench: Benchmarking Image Generation for Multiple Humans (NeurIPS 2025 · similarity 0.87)
- CGS-GAN: 3D Consistent Gaussian Splatting GANs for High Resolution Human Head Synthesis (NeurIPS 2025 · similarity 0.87)
- SyncHuman: Synchronizing 2D and 3D Generative Models for Single-view Human Reconstruction (NeurIPS 2025 · similarity 0.86)
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.
