Research map / Image generation / Editing and style transfer
From Continuity to Editability: Inverting GANs With Consecutive Images
Yangyang Xu, Yong Du, Wenpeng Xiao, Xuemiao Xu, Shengfeng He
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 mutually accessible constraint (linear latent combination) (ablation) | Ours · RAVDESS-12 reconstruction | MSE: 8.065 -> 5.972 | Improved |
| add inversion consistency constraint from consecutive images (ablation) | Ours · RAVDESS-12 reconstruction | MSE: 8.508 -> 5.972 | Improved |
| add mutually accessible constraint (ablation) | Ours · RAVDESS-12 reconstruction | FID: 13.375 -> 13.136 | Improved |
| add inversion consistency constraint (ablation) | Ours · RAVDESS-12 reconstruction | FID: 14.791 -> 13.136 | Improved |
| add mutually accessible constraint (ablation) | Ours · RAVDESS-12 semantic manipulation | FID: 17.274 -> 15.482 | Improved |
| add inversion consistency constraint (ablation) | Ours · RAVDESS-12 semantic manipulation | FID: 16.121 -> 15.482 | Improved |
| add both constraints to Image2StyleGAN baseline (ablation) | Ours · RAVDESS-12 reconstruction | MSE: 8.791 -> 5.972 | Improved |
| add both constraints to Image2StyleGAN baseline (ablation) | Ours · RAVDESS-12 semantic manipulation | NIQE: 3.776 -> 3.254 | Improved |
| invert multiple consecutive images jointly instead of single images | Ours vs InD · RAVDESS-12 reconstruction | MSE: 9.963 -> 5.972 | Improved |
| add noise space N optimization to the inversion | Ours++ vs Ours · RAVDESS-12 reconstruction | MSE: 5.972 -> 0.165 | Improved |
| use consecutive images with other-direction edit on synthesized data | Ours vs I2S · Synthesized dataset semantic manipulation | MSE: 89.671 -> 70.224 | Improved |
| apply to nonlinear StyleFlow edits | Ours vs I2S · Nonlinear dataset semantic edit | MSE: 87.361 -> 69.449 | Improved |
Closest papers by meaning
- RIGID: Recurrent GAN Inversion and Editing of Real Face Videos and Beyond (IJCV 2025 · similarity 0.92)
- RIGID: Recurrent GAN Inversion and Editing of Real Face Videos (ICCV 2023 · similarity 0.92)
- GAN Inversion: A Survey (TPAMI 2022 · similarity 0.90)
- Gradual Residuals Alignment: A Dual-Stream Framework for GAN Inversion and Image Attribute Editing (AAAI 2024 · similarity 0.89)
- In-N-Out: Faithful 3D GAN Inversion with Volumetric Decomposition for Face Editing (CVPR 2024 · similarity 0.88)
- WarpGAN: Warping-Guided 3D GAN Inversion with Style-Based Novel View Inpainting (NeurIPS 2025 · similarity 0.88)
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.
