Research map / Image generation / Editing and style transfer
Implicit Warping for Animation with Image Sets
Arun Mallya, Ting-chun Wang, Ming-Yu 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 residual connection (warped pixels and keys concatenated with queries via MLP) (ablation) | Implicit warping · TalkingHead-1KH 256x256 | FID: 19.978 -> 18.462 | Improved |
| add 400 extra learnable key-value pairs (ablation) | Implicit warping · TalkingHead-1KH 256x256 | FID: 18.462 -> 18.252 | Improved |
| add 400 extra learnable key-value pairs (ablation) | Implicit warping · TalkingHead-1KH 256x256 | PSNR: 23.483 -> 23.582 | Improved |
| use implicit cross-modal attention warping instead of flow-based warping (single source) | Implicit warping vs fv2v · TalkingHead-1KH 512 | FID: 18.09 -> 16.44 | Improved |
| use implicit warping instead of flow-based warping (single source) | Implicit warping vs FOMM · VoxCeleb2 256 | AKD: 1.69 -> 1.75 | Hurt |
| use implicit warping instead of flow-based warping (single source) | Implicit warping vs FOMM · VoxCeleb2 256 | PSNR: 24.15 -> 23.54 | Hurt |
| increase number of source images from 1 to 3 | Implicit warping · TalkingHead-1KH 512 | FID: 15.658 -> 9.097 | Improved |
| increase number of source images from 1 to 3 (average warped features) | FOMM · TalkingHead-1KH 512 | FID: 17.454 -> 19.495 | Hurt |
| increase number of source images from 1 to 3 (average warped features) | face-vid2vid · TalkingHead-1KH 512 | FID: 17.174 -> 19.216 | Hurt |
| increase number of source images from 1 to 3 | Implicit warping · TED Talk 384 | FID: 17.746 -> 13.128 | Improved |
Closest papers by meaning
- Cross-Identity Motion Transfer for Arbitrary Objects through Pose-Attentive Video Reassembling (ECCV 2020 · similarity 0.85)
- One-Shot Flow, Any-Time Frame: A Bidirectional Warping Framework for Event-Based Video Frame Interpolation (CVPR 2026 · similarity 0.85)
- MOWA: Multiple-in-One Image Warping Model (TPAMI 2025 · similarity 0.85)
- MagicAnimate: Temporally Consistent Human Image Animation using Diffusion Model (CVPR 2024 · similarity 0.84)
- Continuous Piecewise-Affine Based Motion Model for Image Animation (AAAI 2024 · similarity 0.84)
- Neural Implicit Morphing of Face Images (CVPR 2024 · similarity 0.83)
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.
