Research map / Image generation / Editing and style transfer
Smoothing the Disentangled Latent Style Space for Unsupervised 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 triplet loss (ablation) | StarGAN v2 · CelebA-HQ gender | FID: 48.35 -> 37.54 | Helped |
| Add triplet loss (ablation) | StarGAN v2 · CelebA-HQ gender | PS: .272 -> .292 | Helped |
| Add style regularization (ablation) | StarGAN v2 · CelebA-HQ gender | PS: .272 -> .432 | Helped |
| Add style regularization and triplet loss (ablation) | StarGAN v2 · CelebA-HQ gender | PS: .272 -> .501 | Helped |
| Add perceptual content loss (ablation) | StarGAN v2 + style regularization + triplet loss · CelebA-HQ gender | FID: 24.29 -> 23.37 | Helped |
| Add perceptual content loss (ablation) | StarGAN v2 + style regularization + triplet loss · CelebA-HQ gender | PS: .501 -> .504 | Helped |
| Add triplet loss alone (ablation) | StarGAN v2 · CelebA-HQ gender | PS: .272 -> .292 | Hurt |
| Replace style regularization with sphere regularization (ablation) | StarGAN v2 + triplet loss · CelebA-HQ gender | PS: .504 -> .897 | Helped |
| Replace style regularization with sphere regularization (ablation) | StarGAN v2 + triplet loss · CelebA-HQ gender | LPIPS: .337 -> .103 | Hurt |
| Add the proposed losses | StarGAN v2 · AFHQ unsupervised | FID: 15.64 -> 14.67 | Helped |
| Add style regularization and triplet loss | TUNIT · AFHQ truly unsupervised | FID: 29.45 -> 16.59 | Helped |
| Add the proposed losses | StarGAN v2 · CelebA-HQ | LPIPS: .442 -> .337; .413 -> .095; .407 -> .128 | Mixed |
Closest papers by meaning
- StEP: Style-Based Encoder Pre-Training for Multi-Modal Image Synthesis (CVPR 2021 · similarity 0.91)
- Homomorphic Latent Space Interpolation for Unpaired Image-To-Image Translation (CVPR 2019 · similarity 0.91)
- Style-Guided and Disentangled Representation for Robust Image-to-Image Translation (AAAI 2022 · similarity 0.90)
- Multi-mapping Image-to-Image Translation via Learning Disentanglement (NeurIPS 2019 · similarity 0.89)
- Domain Adaptive Image-to-Image Translation (CVPR 2020 · similarity 0.89)
- Diffusion-based Image Translation using disentangled style and content representation (ICLR 2023 · similarity 0.89)
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.