Research map / Video / Video generation
Bidirectionally Deformable Motion Modulation For Video-based Human Pose Transfer
Wing-Yin Yu, Lai-Man Po, Ray C.C. Cheung, Yuzhi Zhao, Yu Xue, Kun Li
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 Deformable Motion Modulation (DMM) (ablation) | bidirectional pose transfer model · FashionVideo | FVD-Test128f: 267.476 -> 148.253 | Improved |
| add deformable kernel offsets (DCN) (ablation) | DMM · FashionVideo | LPIPS: 0.0484 -> 0.0478 | Improved |
| add modulated style weight (ablation) | DMM · FashionVideo | LPIPS: 0.0513 -> 0.0478 | Improved |
| add motion mask (ablation) | DMM · FashionVideo | FVD-Test128f: 176.882 -> 148.253 | Improved |
| add forward propagation (ablation) | bidirectional propagation · FashionVideo | FVD-Test128f: 188.869 -> 148.253 | Improved |
| add backward propagation (ablation) | bidirectional propagation · FashionVideo | FVD-Test128f: 156.469 -> 148.253 | Improved |
| add recurrent structural flow (ablation) | bidirectional propagation · FashionVideo | FVD-Test128f: 199.854 -> 148.253 | Improved |
| remove backward propagation (ablation) | bidirectional propagation · FashionVideo | FID: 14.345 -> 14.083 | No clear effect |
| method vs NTED on iPER FID | NTED · iPER | FID: 20.164 -> 22.291 | Hurt |
| proposed method vs GFLA on FashionVideo | GFLA · FashionVideo | FVD-Test128f: 256.430 -> 148.253 | Improved |
Closest papers by meaning
- High-Fidelity Neural Human Motion Transfer From Monocular Video (CVPR 2021 · similarity 0.89)
- Towards High-Quality 3D Motion Transfer with Realistic Apparel Animation (ECCV 2024 · similarity 0.86)
- DLCA-Recon: Dynamic Loose Clothing Avatar Reconstruction from Monocular Videos (AAAI 2024 · similarity 0.85)
- D^3-Human: Dynamic Disentangled Digital Human from Monocular Video (CVPR 2025 · similarity 0.84)
- HumanMM: Global Human Motion Recovery from Multi-shot Videos (CVPR 2025 · similarity 0.84)
- Neural Pose Representation Learning for Generating and Transferring Non-Rigid Object Poses (NeurIPS 2024 · 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.
