Research map / Video / Video segmentation and tracking
DINet: Deformation Inpainting Network for Realistic Face Visually Dubbing on High Resolution Video
Zhimeng Zhang, Zhipeng Hu, Wenjin Deng, Changjie Fan, Tangjie Lv, Yu Ding
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 |
|---|---|---|---|
| Use DINet instead of the strongest reported baseline on visual quality | Face visually dubbing · HDTF | SSIM: 0.9078 -> 0.9425 | Improved |
| Use DINet instead of the strongest reported baseline on visual quality | Face visually dubbing · HDTF | LPIPS: 0.0576 -> 0.0289 | Improved |
| Use DINet instead of the strongest reported baseline on visual quality | Face visually dubbing · MEAD-Neutral | SSIM: 0.8994 -> 0.9204 | Improved |
| Use DINet instead of the strongest reported baseline on visual quality | Face visually dubbing · MEAD-Neutral | LPIPS: 0.0779 -> 0.0459 | Improved |
| Use DINet instead of the strongest reported baseline | Face visually dubbing · HDTF | LSE-D: 6.6137 -> 8.3771 | Hurt |
| Use DINet instead of the strongest reported baseline | Face visually dubbing · MEAD-Neutral | LSE-D: 6.0218 -> 7.5157 | Hurt |
| Add AdaAT deformation rather than removing it and using AdaIN feature injection (ablation) | DINet · HDTF | SSIM: 0.9147 -> 0.9425 | Improved |
| Add source-feature fusion through the inpainting part (ablation) | DINet · HDTF | SSIM: 0.8691 -> 0.9425 | Improved |
| Use AdaAT deformation instead of attention-based feature selection (ablation) | DINet · HDTF | LPIPS: 0.0433 -> 0.0289 | Improved |
| Use AdaAT deformation instead of dense flow (ablation) | DINet · HDTF | LPIPS: 0.0656 -> 0.0289 | Improved |
| Use AdaAT rather than dense-flow deformation on misaligned reference features (ablation) | DINet · HDTF | PSNR: 23.5656 -> 30.0082 | Improved |
Later papers that compared with this method
1 later accepted paper reports DINet as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Identity-Preserving Video Dubbing Using Motion Warping (IJCV 2026 · better on 10 of 10 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
- StyleHEAT: One-Shot High-Resolution Editable Talking Face Generation via Pre-trained StyleGAN (ECCV 2022 · similarity 0.86)
- Learning Dynamic Facial Radiance Fields for Few-Shot Talking Head Synthesis (ECCV 2022 · similarity 0.86)
- Synergizing Motion and Appearance: Multi-Scale Compensatory Codebooks for Talking Head Video Generation (CVPR 2025 · similarity 0.85)
- Enhancing Identity-Deformation Disentanglement in StyleGAN for One-Shot Face Video Re-Enactment (AAAI 2025 · similarity 0.85)
- Occlusion-Insensitive Talking Head Video Generation via Facelet Compensation (AAAI 2025 · similarity 0.85)
- RealPortrait: Realistic Portrait Animation with Diffusion Transformers (AAAI 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.
