Research map / Image generation / Text-to-image diffusion
FairHuman: Boosting Hand and Face Quality in Human Image Generation with Minimum Potential Delay Fairness in Diffusion Models
Yuxuan Wang, Tianwei Cao, Huayu Zhang, Zhongjiang He, Kongming Liang, Zhanyu Ma
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 hand and face region losses to the global diffusion loss (fixed weights) (ablation) | FairHuman LoRA · Human validation set (LoRA) | FID: 49.48 -> 47.22 | Improved |
| add MPD fairness weighting instead of fixed-weight scalarization (ablation) | FairHuman LoRA · Human validation set (LoRA) | FID: 47.22 -> 46.11 | Improved |
| add regional losses to ControlNet (ablation) | FairHuman ControlNet · Human validation set (ControlNet) | pose mean difference: 3.45 -> 2.78 | Improved |
| add regional losses without fairness to ControlNet (ablation) | FairHuman ControlNet · Human validation set (ControlNet) | FID: 46.18 -> 46.79 | Hurt |
| add MPD fairness weighting to ControlNet multi-objective training (ablation) | FairHuman ControlNet · Human validation set (ControlNet) | pose mean difference: 2.78 -> 2.35 | Improved |
| increase LoRA rank from 64 to 256 (ablation) | FairHuman LoRA · Human validation set | FID: 47.87 -> 46.11 | Improved |
| use MPD fairness instead of other multi-objective strategies (DWA, UW, RLW, SI, Nash-MTL, LS) (ablation) | FairHuman LoRA · Human validation set | HPS (%): 32.73 vs 32.52 to 32.63 | Mixed |
| use single fairness-weighted LoRA instead of MoLE two-LoRA mixture | FairHuman LoRA · SDXL backbone | extra memory usage: 2611.75MB -> 1413.12MB | Improved |
| apply method to ControlNet instead of using ControlNet-Union | ControlNet · Human validation set | HPS (%): 32.80 vs 32.72 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- MoLE — MoLE: Enhancing Human-centric Text-to-image Diffusion via Mixture of Low-rank Experts (NeurIPS 2024) · better on 7 of 7 shared results
- SD-XL Base — SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis (ICLR 2024) · better on 3 of 3 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
- Fair Text-to-Image Diffusion via Fair Mapping (AAAI 2025 · similarity 0.87)
- HyperHuman: Hyper-Realistic Human Generation with Latent Structural Diffusion (ICLR 2024 · similarity 0.87)
- MoLE: Enhancing Human-centric Text-to-image Diffusion via Mixture of Low-rank Experts (NeurIPS 2024 · similarity 0.87)
- MultiHuman-Testbench: Benchmarking Image Generation for Multiple Humans (NeurIPS 2025 · similarity 0.86)
- HanDiffuser: Text-to-Image Generation With Realistic Hand Appearances (CVPR 2024 · similarity 0.86)
- CapHuman: Capture Your Moments in Parallel Universes (CVPR 2024 · similarity 0.85)
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.
