Research map / 3D vision / Human pose estimation
Liquid Warping GAN With Attention: A Unified Framework for Human Image Synthesis
Wen Liu, Zhixin Piao, Zhi Tu, Wenhan Luo, Lin Ma, Shenghua Gao
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 personalization (one/few-shot finetuning) (ablation) | AttLWB · Youtube-Dancer-18 | PSNR: 16.932 -> 17.974 | Improved |
| add personalization (one/few-shot finetuning) (ablation) | AttLWB · Youtube-Dancer-18 | FID: 79.321 -> 59.832 | Improved |
| replace LWB with attentional AttLWB (ablation) | LWB · Novel view synthesis (MotionSynthetic) | PSNR: 25.055 -> 28.625 | Improved |
| replace LWB with attentional AttLWB (ablation) | LWB · One-shot motion imitation self-imitation | Face-CS: 0.560 -> 0.591 | Improved |
| replace LWB with attentional AttLWB (ablation) | LWB · Appearance transfer (iPER) | SSIM: 0.734 -> 0.726 | Hurt |
| add adversarial loss to perceptual loss (ablation) | AttLWB · Youtube-Dancer-18 | PSNR: 18.204 -> 19.656 | Improved |
| add face identity loss to perceptual + adversarial (ablation) | AttLWB · Youtube-Dancer-18 | PSNR: 19.656 -> 19.542 | Mixed |
| add attention regularization loss to complete the full loss (ablation) | AttLWB · Youtube-Dancer-18 | PSNR: 19.542 -> 20.038 | Improved |
| increase number of source images s_n with small steps T=10 (ablation) | AttLWB · Youtube-Dancer-18 | PSNR: s_n=8, T=10 decreases | Hurt |
| increase number of source images s_n with large T (ablation) | AttLWB · Youtube-Dancer-18 | PSNR | Improved |
| increase personalization steps T beyond about 150 (ablation) | AttLWB · Youtube-Dancer-18 | PSNR: saturates at around 150 steps | No clear effect |
| use attention aggregation (AttLWB) over other warping strategies (ablation) | AttLWB · User study (64 volunteers) | user preference %: 41.73% vs 20.04% | Improved |
Closest papers by meaning
- Liquid Warping GAN: A Unified Framework for Human Motion Imitation, Appearance Transfer and Novel View Synthesis (ICCV 2019 · similarity 0.97)
- Soft-Gated Warping-GAN for Pose-Guided Person Image Synthesis (NeurIPS 2018 · similarity 0.91)
- BodyGAN: General-Purpose Controllable Neural Human Body Generation (CVPR 2022 · similarity 0.88)
- FW-GAN: Flow-Navigated Warping GAN for Video Virtual Try-On (ICCV 2019 · similarity 0.87)
- CGS-GAN: 3D Consistent Gaussian Splatting GANs for High Resolution Human Head Synthesis (NeurIPS 2025 · similarity 0.87)
- CapHuman: Capture Your Moments in Parallel Universes (CVPR 2024 · 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.
