Research map / Image generation / Autoregressive and latent generation
High-Resolution GAN Inversion for Degraded Images in Large Diverse Datasets
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 CRI instead of PTI | StyleGAN-XL inversion · ImageNet val inpainting | LPIPS: 0.1219 -> 0.1164 | Helped |
| use CRI instead of PTI | StyleGAN-XL inversion · ImageNet val colorization | FID: 47.51 -> 45.90 | Helped |
| use CRI instead of PTI | StyleGAN-XL inversion · ImageNet val SR | NIQE: 5.5659 -> 5.3065 | Helped |
| use CRI instead of the strongest baseline | StyleGAN2 inversion · CelebA-HQ inpainting | FID: 53.48 -> 48.00 | Helped |
| use CRI instead of the strongest baseline | StyleGAN2 inversion · CelebA-HQ colorization | LPIPS: 0.1621 -> 0.1505 | Helped |
| use CRI instead of the strongest baseline | StyleGAN2 inversion · CelebA-HQ SR | FID: 60.76 -> 58.88 | Helped |
| use CRI with StyleGAN-XL instead of DGP with BigGAN | ImageNet restoration · ImageNet colorization | FID: 189.42 -> 92.48 | Helped |
| add L2 regularized offset rather than no regularization (ablation) | StyleGAN2 CRI · CelebA-HQ colorization | LPIPS: 0.1996 -> 0.1560 | Helped |
| add L2 regularized offset rather than no regularization (ablation) | StyleGAN2 CRI · CelebA-HQ colorization | ID Similarity: 0.7251 -> 0.5756 | Hurt |
| replace L2 offset regularization with L1 (ablation) | StyleGAN2 CRI · CelebA-HQ colorization | LPIPS: 0.1560 -> 0.1694 | Hurt |
| replace L2 offset regularization with L1 (ablation) | StyleGAN2 CRI · CelebA-HQ colorization | ID Similarity: 0.5756 -> 0.1955 | Hurt |
| increase the number of clustering centroids (ablation) | CRI · ImageNet colorization | qualitative performance | Helped |
Closest papers by meaning
- Robust Unsupervised StyleGAN Image Restoration (CVPR 2023 · similarity 0.92)
- HyperInverter: Improving StyleGAN Inversion via Hypernetwork (CVPR 2022 · similarity 0.90)
- ReGANIE: Rectifying GAN Inversion Errors for Accurate Real Image Editing (AAAI 2023 · similarity 0.90)
- WarpGAN: Warping-Guided 3D GAN Inversion with Style-Based Novel View Inpainting (NeurIPS 2025 · similarity 0.89)
- Gradual Residuals Alignment: A Dual-Stream Framework for GAN Inversion and Image Attribute Editing (AAAI 2024 · similarity 0.87)
- Style Quantization for Data-Efficient GAN Training (CVPR 2025 · similarity 0.86)
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.