Research map / Image generation / GANs
CoordGAN: Self-Supervised Dense Correspondences Emerge From GANs
Check 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 |
|---|---|---|---|
| use CoordGAN correspondence maps for label propagation | dense correspondence evaluation · CelebA-HQ | IoU: 48.50 -> 52.25 | Helped |
| use CoordGAN correspondence maps for label propagation | dense correspondence evaluation · DGAN-face | IoU: 20.36 -> 23.78 | Helped |
| use CoordGAN correspondence maps for label propagation | dense correspondence evaluation · DGAN-car | IoU: 11.07 -> 13.23 | Helped |
| represent structure with a shared coordinate map and vary texture codes | CoordGAN texture swapping · CelebA-HQ, Stanford Cars, AFHQ-cat | LPIPS: 0.58 -> 0.22; 0.61 -> 0.21; 0.55 -> 0.27 | Helped |
| modulate multiple generator layers with the correspondence map (ablation) | CoordGAN · CelebA-HQ | IoU: 48.59 -> 52.25 | Helped |
| add the warp loss (ablation) | CoordGAN generator · CelebA-HQ | IoU: 24.84 -> 52.25 | Helped |
| add the warp loss (ablation) | CoordGAN generator · DGAN-face | IoU: 10.52 -> 23.78 | Helped |
| add the structure swapping constraint (ablation) | CoordGAN generator · CelebA-HQ | IoU: 46.52 -> 52.25 | Helped |
| add the texture swapping constraint (ablation) | CoordGAN generator · CelebA-HQ | IoU: 45.96 -> 52.25 | Helped |
| add a backward MLP with an L1 coordinate supervision loss (ablation) | CoordGAN · CelebA-HQ | IoU: 52.25 -> 54.51 | Helped |
| add a backward MLP with an L1 coordinate supervision loss (ablation) | CoordGAN · DGAN-face | IoU: 23.78 -> 25.44 | Helped |
| add a backward MLP with an L1 coordinate supervision loss (ablation) | CoordGAN · DGAN-car | IoU: 13.23 -> 12.59 | Hurt |
Closest papers by meaning
- GAN-Supervised Dense Visual Alignment (CVPR 2022 · similarity 0.90)
- Disentangled3D: Learning a 3D Generative Model With Disentangled Geometry and Appearance From Monocular Images (CVPR 2022 · similarity 0.88)
- A 3D GAN for Improved Large-Pose Facial Recognition (CVPR 2021 · similarity 0.87)
- A Geometric Analysis of Deep Generative Image Models and Its Applications (ICLR 2021 · similarity 0.87)
- MOST-GAN: 3D Morphable StyleGAN for Disentangled Face Image Manipulation (AAAI 2022 · similarity 0.87)
- Learning Dense Correspondence via 3D-Guided Cycle Consistency (CVPR 2016 · similarity 0.86)
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.