Research map / Video / Video generation
Cross-Identity Motion Transfer for Arbitrary Objects through Pose-Attentive Video Reassembling
Subin Jeon, Seonghyeon Nam, Seoung Wug Oh, Seon Joo Kim
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 cross-identity reconstruction training (ablation) | pose-attentive reassembling network · VoxCeleb2 (cross-identity) | FID: 55.11 -> 36.39 | Improved |
| add cross-identity reconstruction training (ablation) | pose-attentive reassembling network · VoxCeleb2 (cross-identity) | AED: 0.76 -> 0.62 | Improved |
| add cross-identity reconstruction training (ablation) | pose-attentive reassembling network · VoxCeleb2 (self-reconstruction) | AKD: 5.79 -> 1.90 | Improved |
| add cross-identity reconstruction training (ablation) | pose-attentive reassembling network · VoxCeleb2 (self-reconstruction) | FID: 18.49 -> 17.24 | Improved |
| add cross-identity reconstruction training (ablation) | pose-attentive reassembling network · VoxCeleb2 (self-reconstruction) | AED: 0.39 -> 0.38 | No clear effect |
| use pose-attentive retrieval instead of ConcatAppearance (ablation) | pose-attentive reassembling network · VoxCeleb2 (cross-identity) | FID: 63.52 -> 36.39 | Improved |
| use pose-attentive retrieval instead of ConcatAppearance (ablation) | pose-attentive reassembling network · VoxCeleb2 (self-reconstruction) | AKD: 4.21 -> 1.90 | Improved |
| use pose-attentive retrieval instead of ConcatNearest (ablation) | pose-attentive reassembling network · VoxCeleb2 (cross-identity) | FID: 52.13 -> 36.39 | Improved |
| use pose-attentive retrieval instead of ConcatAverage (ablation) | pose-attentive reassembling network · VoxCeleb2 (self-reconstruction) | AKD: 103.78 -> 1.90 | Improved |
| increase source frames from 1 to 32 | pose-attentive reassembling network · VoxCeleb2 few-shot | AED: 0.46 -> 0.29 | Improved |
| increase source frames from 1 to 32 | pose-attentive reassembling network · VoxCeleb2 few-shot | AKD: 7.47 -> 4.09 | Improved |
| increase source frames from 1 to 32 | pose-attentive reassembling network · VoxCeleb2 few-shot | FID: 56.19 -> 60.80 | Hurt |
Closest papers by meaning
- Motion and Appearance Adaptation for Cross-Domain Motion Transfer (ECCV 2022 · similarity 0.86)
- Repurposing Video Diffusion Transformers for Cross-View Temporal Object Correspondence (NeurIPS 2026 · similarity 0.86)
- Large Displacement Motion Transfer with Unsupervised Anytime Interpolation (ICML 2025 · similarity 0.86)
- High-Fidelity Neural Human Motion Transfer From Monocular Video (CVPR 2021 · similarity 0.86)
- X-NeMo: Expressive Neural Motion Reenactment via Disentangled Latent Attention (ICLR 2025 · similarity 0.85)
- DisPose: Disentangling Pose Guidance for Controllable Human Image Animation (ICLR 2025 · 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.
