Research map / Image generation / Editing and style transfer
Flow-based Image-to-Image Translation with Feature Disentanglement
Ruho Kondo, Keisuke Kawano, Satoshi Koide, Takuro Kutsuna
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 |
|---|---|---|---|
| Lower sampling temperature from T=1.0 to 0.8 | FUNS · CelebA | FID: 39.6 ± 3.9 -> 29.5 ± 3.5 | Improved |
| Lower sampling temperature from T=1.0 to 0.8 | FUNS · CHC | FID: 10.5 ± 2.0 -> 11.1 ± 2.5 | No clear effect |
| Lower sampling temperature from T=1.0 to 0.8 | VUNet · CelebA | FID: 66.0 ± 4.3 -> 81.7 ± 3.7 | Hurt |
| Lower sampling temperature from T=1.0 to 0.8 | VUNet · CHC | FID: 96.5 ± 2.1 -> 164.0 ± 5.7 | Hurt |
| Lower sampling temperature from T=1.0 to 0.8 | VUNet · CelebA | LPIPS: 0.148 -> 0.105 | Hurt |
| Use FUNS instead of VUNet (T=1.0) | image-to-image translation · CelebA | FID: 66.0 ± 4.3 -> 39.6 ± 3.9 | Improved |
| Use FUNS instead of VUNet (T=1.0) | image-to-image translation · CHC | FID: 96.5 ± 2.1 -> 10.5 ± 2.0 | Improved |
| Use FUNS instead of PUNet (T=1.0) | image-to-image translation · CelebA | c-LPIPS: 0.180 -> 0.262 | Improved |
| Use FUNS instead of VUNet (T=1.0) | image-to-image translation · CHC | c-LPIPS: 0.118 -> 0.157 | Improved |
| Use FUNS instead of VUNet for condition fidelity | image-to-image translation · CelebA | Smile classification accuracy: 0.977 -> 0.973 | Hurt |
| Use FUNS instead of VUNet for condition fidelity | image-to-image translation · CHC | L2 error: 9.46 -> 9.67 | Hurt |
| Measure c-LPIPS instead of LPIPS | diversity metric · CHC | diversity: 0.217 vs 0.118 | Mixed |
Closest papers by meaning
- VFlow: More Expressive Generative Flows with Variational Data Augmentation (ICML 2020 · similarity 0.86)
- Phase-wise Velocity Distillation: Towards Effective Image Generation with A Single NFE (NeurIPS 2026 · similarity 0.86)
- Flow Matching with General Discrete Paths: A Kinetic-Optimal Perspective (ICLR 2025 · similarity 0.85)
- Equilibrium Matching: Generative Modeling with Implicit Energy-Based Models (NeurIPS 2026 · similarity 0.85)
- Image Inpainting via Tractable Steering of Diffusion Models (ICLR 2024 · similarity 0.85)
- Align Your Flow: Scaling Continuous-Time Flow Map Distillation (NeurIPS 2025 · similarity 0.85)
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.
