Research map / Image generation / Editing and style transfer
Few-Shot Unsupervised Image-to-Image Translation
Ming-Yu Liu, Xun Huang, Arun Mallya, Tero Karras, Timo Aila, Jaakko Lehtinen, Jan Kautz
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 |
|---|---|---|---|
| increase test-time class images from 1 to 5 | FUNIT · Animal Faces (unseen classes) | Top5-test accuracy: 82.36 -> 96.05 | Improved |
| increase test-time class images from 1 to 5 | FUNIT · Animal Faces (unseen classes) | mFID: 93.03 -> 70.24 | Improved |
| increase test-time class images from 5 to 20 | FUNIT · Animal Faces (unseen classes) | Top5-test accuracy: 96.05 -> 97.96 | Improved |
| add the zero-centered gradient penalty (ablation) | FUNIT-1 · Animal Faces, FUNIT-1 ablation | Top5-test accuracy: 15.50 -> 82.36 | Improved |
| add the feature matching loss (ablation) | FUNIT-1 · Animal Faces, FUNIT-1 ablation | Top1-all accuracy: 15.33 -> 17.07 | Improved |
| add the feature matching loss (ablation) | FUNIT-1 · Animal Faces, FUNIT-1 ablation | Top5-test accuracy: 82.43 -> 82.36 | No clear effect |
| increase content reconstruction weight lambda_R from 0.1 to 10 (ablation) | FUNIT-1 · Animal Faces, FUNIT-1 ablation | Top5-test accuracy: 82.36 -> 75.81 | Hurt |
| increase content reconstruction weight lambda_R from 0.1 to 10 (ablation) | FUNIT-1 · Animal Faces, FUNIT-1 ablation | DIPD: 1.364 -> 1.298 | Improved |
| raise lambda_R from 0.01 to 0.1 (ablation) | FUNIT-1 · Animal Faces, FUNIT-1 ablation | Top5-test accuracy: 81.68 -> 82.36 | Improved |
| train FUNIT with more source classes (69 to 119) | FUNIT-1 · Animal Faces (unseen classes) | translation accuracy, IS, mFID | Improved |
| use generated FUNIT images (N=100) for few-shot classification | classifier · Animal Faces few-shot classification (1-shot, avg of 5 splits) | accuracy: 38.76 -> 42.14 | Improved |
| use FUNIT images instead of S&H features (N=100) | classifier · Animal Faces few-shot classification (1-shot, avg of 5 splits) | accuracy: 40.76 -> 42.14 | Improved |
Closest papers by meaning
- Unsupervised Image-to-Image Translation Networks (NeurIPS 2017 · similarity 0.93)
- Unsupervised Image-to-Image Translation With Generative Prior (CVPR 2022 · similarity 0.91)
- TraVeLGAN: Image-To-Image Translation by Transformation Vector Learning (CVPR 2019 · similarity 0.91)
- Unpaired Image-To-Image Translation Using Cycle-Consistent Adversarial Networks (ICCV 2017 · similarity 0.91)
- Unsupervised Attention-guided Image-to-Image Translation (NeurIPS 2018 · similarity 0.90)
- COCO-FUNIT: Few-Shot Unsupervised Image Translation with a Content Conditioned Style Encoder (ECCV 2020 · similarity 0.90)
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.
