Research map / Image generation / Editing and style transfer
High-Fidelity Image Inpainting with GAN Inversion
Yongsheng Yu, Libo Zhang, Heng Fan, Tiejian Luo
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 soft-update mean latent (SML) (ablation) | InvertFill · Places2 val (extreme masks, ablation) | FID: 35.37 -> 24.73 | Improved |
| add pre-modulation (PM) (ablation) | InvertFill · Places2 val (extreme masks, ablation) | FID: 24.73 -> 21.08 | Improved |
| add pre-modulation (PM) (ablation) | InvertFill · Places2 val (extreme masks, ablation) | SSIM: 0.358 -> 0.366 | Improved |
| add pre-modulation (PM) (ablation) | InvertFill · Places2 val (extreme masks, ablation) | PSNR: 13.99 -> 14.62 | Improved |
| add soft-update mean latent (SML) (ablation) | InvertFill · Places2 val (extreme masks, ablation) | LPIPS: 0.395 -> 0.389 | No clear effect |
| use F&W+ latents instead of only W+ (ablation) | InvertFill · Places2 val (extreme masks, ablation) | FID: 42.85 -> 21.08 | Improved |
| replace pSp+Blend with InvertFill | pSp + Blend · CelebA-HQ (extreme masks, vs inversion/diffusion) | FID: 21.96 -> 13.21 | Improved |
| add image blending post-processing to pSp | pSp · CelebA-HQ (extreme masks, vs inversion/diffusion) | FID: 25.61 -> 21.96 | Improved |
| replace Score-SDE with InvertFill | Score-SDE · CelebA-HQ (extreme masks, vs inversion/diffusion) | FID: 24.76 -> 13.21 | Improved |
| use InvertFill instead of ProFill on extreme masks | ProFill · Places2 val (extreme masks) | SSIM: 0.409 -> 0.366 | Hurt |
| use InvertFill instead of ProFill on extreme masks | ProFill · Places2 val (extreme masks) | FID: 46.44 -> 21.08 | Improved |
| apply InvertFill to outpainting without retraining generator | NS-outpaint · Scenery outpainting | FID: 38.95 -> 20.90 | Improved |
Closest papers by meaning
- High-Fidelity Image Inpainting with Multimodal Guided GAN Inversion (IJCV 2025 · similarity 0.97)
- Diverse Inpainting and Editing with GAN Inversion (ICCV 2023 · similarity 0.94)
- Dual-Path Image Inpainting With Auxiliary GAN Inversion (CVPR 2022 · similarity 0.92)
- WarpGAN: Warping-Guided 3D GAN Inversion with Style-Based Novel View Inpainting (NeurIPS 2025 · similarity 0.92)
- InOut: Diverse Image Outpainting via GAN Inversion (CVPR 2022 · similarity 0.92)
- In-Domain GAN Inversion for Faithful Reconstruction and Editability (TPAMI 2024 · similarity 0.91)
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.
