Research map / Image generation / Editing and style transfer
Encoding in Style: A StyleGAN Encoder for Image-to-Image Translation
Check 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 |
|---|---|---|---|
| Add the identity loss (ablation) | pSp · CelebA-HQ image reconstruction | Similarity: 0.19 -> 0.56 | Helped |
| Add the identity loss (ablation) | pSp · CelebA-HQ image reconstruction | LPIPS: 0.17 -> 0.17 | No clear effect |
| Add the identity loss (ablation) | pSp · CelebA-HQ image reconstruction | MSE: 0.03 -> 0.03 | No clear effect |
| Use naive W+ prediction instead of a single W vector (ablation) | StyleGAN inversion encoder · CelebA-HQ image reconstruction | Similarity: 0.35 -> 0.49 | Helped |
| Use the hierarchical feature-pyramid pSp encoder instead of naive W+ prediction (ablation) | StyleGAN inversion encoder · CelebA-HQ image reconstruction | Similarity: 0.49 -> 0.56 | Helped |
| Use pSp instead of R&R | Face frontalization · FEI Face Database, 90° input rotation | Similarity: 0.34 -> 0.32 | Hurt |
| Use pSp instead of R&R | Face frontalization · FEI Face Database, 70° input rotation | Similarity: 0.56 -> 0.52 | Hurt |
| Use pSp instead of R&R | Face frontalization · FEI Face Database, 50° input rotation | Similarity: 0.66 -> 0.60 | Hurt |
| Use pSp instead of R&R | Face frontalization · FEI Face Database, 30° input rotation | Similarity: 0.7 -> 0.63 | Hurt |
| Train an encoder against a fixed pretrained StyleGAN generator rather than train an adversary for translation | pSp · Face frontalization | Runtime: 1.5s -> 0.1s | Helped |
| Use pSp instead of pix2pixHD | Conditional image synthesis · CelebA-HQ segmentation | Human preference for pSp: 94.72% | Helped |
| Use pSp instead of SPADE | Conditional image synthesis · CelebA-HQ segmentation | Human preference for pSp: 95.25% | Helped |
Closest papers by meaning
- StEP: Style-Based Encoder Pre-Training for Multi-Modal Image Synthesis (CVPR 2021 · similarity 0.91)
- Masked and Adaptive Transformer for Exemplar Based Image Translation (CVPR 2023 · similarity 0.89)
- A Style-Aware Discriminator for Controllable Image Translation (CVPR 2022 · similarity 0.89)
- Memory-Guided Unsupervised Image-to-Image Translation (CVPR 2021 · similarity 0.88)
- HyperStyle: StyleGAN Inversion With HyperNetworks for Real Image Editing (CVPR 2022 · similarity 0.88)
- Image-To-Image Translation via Group-Wise Deep Whitening-And-Coloring Transformation (CVPR 2019 · 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.