Research map / Image generation / Editing and style transfer
Improving Shape Deformation in Unsupervised Image-to-Image Translation
Aaron Gokaslan, Vivek Ramanujan, Daniel Ritchie, Kwang In Kim, James Tompkin
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 |
|---|---|---|---|
| add dilated discriminator instead of patch discriminator (ablation) | GANimorph · Cat->Dog | target class pixels (%): 42.90 -> 98.57 | Improved |
| add dilated discriminator instead of patch discriminator (ablation) | GANimorph · Dog->Cat | target class pixels (%): 91.77 -> 100.00 | Improved |
| add dilated discriminator instead of fully connected discriminator (ablation) | GANimorph · Cat->Dog | target class pixels (%): 93.60 -> 98.57 | Improved |
| add dilated discriminator instead of fully connected discriminator (ablation) | GANimorph · Dog->Cat | target class pixels (%): 91.41 -> 100.00 | Improved |
| add feature matching loss (ablation) | GANimorph · Cat->Dog | target class pixels (%): 93.64 -> 98.57 | Improved |
| add feature matching loss (ablation) | GANimorph · Dog->Cat | target class pixels (%): 85.62 -> 100.00 | Improved |
| add MS-SSIM perceptual cyclic loss instead of L1 only (ablation) | GANimorph · Cat->Dog | target class pixels (%): 0.00 -> 98.57 | Improved |
| add MS-SSIM perceptual cyclic loss instead of L1 only (ablation) | GANimorph · Dog->Cat | target class pixels (%): 0.00 -> 100.00 | Improved |
| use GANimorph generator with fully connected discriminator instead of DiscoGAN | DiscoGAN · Cat->Dog | target class pixels (%): 75.38 -> 93.60 | Improved |
| use GANimorph instead of CycleGAN | CycleGAN · Cat->Dog | target class pixels (%): 0.01 -> 98.57 | Improved |
Closest papers by meaning
- TransGaGa: Geometry-Aware Unsupervised Image-To-Image Translation (CVPR 2019 · similarity 0.88)
- Deforming Autoencoders: Unsupervised Disentangling of Shape and Appearance (ECCV 2018 · similarity 0.88)
- UNIST: Unpaired Neural Implicit Shape Translation Network (CVPR 2022 · similarity 0.87)
- Semantic-Aware Implicit Template Learning via Part Deformation Consistency (ICCV 2023 · similarity 0.87)
- Unsupervised Learning of Dense Shape Correspondence (CVPR 2019 · similarity 0.87)
- Learning Category-Specific Mesh Reconstruction from Image Collections (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.
