Research map / Image generation / Editing and style transfer
Unpaired Image-to-Image Translation With Shortest Path Regularization
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 |
|---|---|---|---|
| add path length regularization (Lpath) (ablation) | shortest path unpaired translation · label→city | mAP: 28.3 -> 31.0 | Helped |
| add path length regularization (Lpath) (ablation) | shortest path unpaired translation · old→young | FID: 44.0 -> 41.9 | Helped |
| add path length regularization (Lpath) (ablation) | interpolations on old→young · old→young interpolation | interpolation FID: 41.27 -> 28.62 | Helped |
| add reconstruction loss Lrec and KL loss on top of GAN+identity (ablation) | shortest path unpaired translation · label→city | mAP: 23.1 -> 26.0 -> 28.3 | Helped |
| replace path regularization with Euclidean cross-domain distance (ablation) | shortest path unpaired translation · label→city | mAP: 31.0 -> 29.5 | Hurt |
| replace path regularization with VGG cross-domain distance (ablation) | shortest path unpaired translation · label→city | mAP: 31.0 -> 22.8 | Hurt |
| use the proposed method (shortest path regularization) vs baselines | unpaired image translation · cat→dog | FID: 46.1 vs 46.4 (SRC) and 48.5 (NEGCUT) | Helped |
| use the proposed method vs baselines | unpaired image translation · horse→zebra | FID: 52.1 vs 34.4 (SRC, best) | Hurt |
| use the proposed method vs baselines | unpaired image translation · winter→summer | FID: 70.9 vs 71.6 (SRC) | Helped |
| add intermediate-domain path length regularization on multi-layer features instead of final-domain distance (ablation) | shortest path unpaired translation · label→city | mAP: 31.0 (full) vs 29.5 (Euclidean variant) | Mixed |
Closest papers by meaning
- Unsupervised Image-to-Image Translation Using Domain-Specific Variational Information Bound (NeurIPS 2018 · similarity 0.88)
- Unsupervised Image-to-Image Translation with Density Changing Regularization (NeurIPS 2022 · similarity 0.87)
- Extremal Domain Translation with Neural Optimal Transport (NeurIPS 2023 · similarity 0.87)
- Homomorphic Latent Space Interpolation for Unpaired Image-To-Image Translation (CVPR 2019 · similarity 0.87)
- Unsupervised Image-to-Image Translation Networks (NeurIPS 2017 · similarity 0.86)
- Unpaired Image-to-Image Translation via Neural Schrödinger Bridge (ICLR 2024 · 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.