Research map / Image generation / Editing and style transfer
TransGaGa: Geometry-Aware Unsupervised Image-To-Image Translation
Wayne Wu, Kaidi Cao, Cheng Li, Chen Qian, Chen Change Loy
Check 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 |
|---|---|---|---|
| Use geometry-aware disentangled translation instead of the strongest baseline | Face translation · Face translation average | FID: 50.64 -> 37.73 | Improved |
| Use geometry-aware disentangled translation instead of the strongest baseline | Face translation · Face translation average | Diversity (LPIPS): 0.46 -> 0.54 | Improved |
| Add appearance and geometry transformers (ablation) | TransGaGa · Horse → giraffe and human → cat face | Realism (testers labeled real): 0.0% -> 16.2%; 0.0% -> 23.9% | Improved |
| Add cycle-consistency loss (ablation) | TransGaGa · Horse → giraffe and human → cat face | Realism (testers labeled real): 14.2% -> 16.2%; 16.8% -> 23.9% | Improved |
| Add KL loss (ablation) | TransGaGa · Horse → giraffe and human → cat face | Realism (testers labeled real): 10.2% -> 16.2%; 15.4% -> 23.9% | Improved |
| Use VGG loss instead of L1 loss (ablation) | TransGaGa · Horse → giraffe and human → cat face | Realism (testers labeled real): 12.6% -> 16.2%; 18.4% -> 23.9% | Improved |
Closest papers by meaning
- Improving Shape Deformation in Unsupervised Image-to-Image Translation (ECCV 2018 · similarity 0.88)
- Model-based occlusion disentanglement for image-to-image translation (ECCV 2020 · similarity 0.88)
- TraVeLGAN: Image-To-Image Translation by Transformation Vector Learning (CVPR 2019 · similarity 0.87)
- Towards Instance-Level Image-To-Image Translation (CVPR 2019 · similarity 0.87)
- Multi-mapping Image-to-Image Translation via Learning Disentanglement (NeurIPS 2019 · similarity 0.87)
- Multimodal Unsupervised Image-to-image Translation (ECCV 2018 · similarity 0.87)
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.
