Research map / Image generation / Editing and style transfer
Disentangled and Controllable Face Image Generation via 3D Imitative-Contrastive Learning
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 |
|---|---|---|---|
| add imitative losses (ablation) | StyleGAN with adversarial loss · FFHQ | DS_α: 0.83 -> 13.4 | Helped |
| add imitative losses (ablation) | StyleGAN with adversarial loss · FFHQ | DS_β: 1.98 -> 37.0 | Helped |
| add imitative losses (ablation) | StyleGAN with adversarial loss · FFHQ | DS_γ: 0.87 -> 40.4 | Helped |
| add imitative losses (ablation) | StyleGAN with adversarial loss · FFHQ | DS_θ: 0.07 -> 31.6 | Helped |
| add imitative losses (ablation) | StyleGAN with adversarial loss · FFHQ | FID: 5.49 -> 9.15 | Hurt |
| add imitative losses (ablation) | StyleGAN with adversarial loss · FFHQ | PPL: 106 -> 102 | No clear effect |
| add contrastive losses to imitative training (ablation) | DiscoFaceGAN · FFHQ | DS_α: 13.4 -> 7.85 | Hurt |
| add contrastive losses to imitative training (ablation) | DiscoFaceGAN · FFHQ | DS_β: 37.0 -> 80.4 | Helped |
| add contrastive losses to imitative training (ablation) | DiscoFaceGAN · FFHQ | DS_γ: 40.4 -> 489 | Helped |
| add contrastive losses to imitative training (ablation) | DiscoFaceGAN · FFHQ | DS_θ: 31.6 -> 36.7 | Helped |
| add contrastive losses to imitative training (ablation) | DiscoFaceGAN · FFHQ | FID: 9.15 -> 12.9 | Hurt |
| add contrastive losses to imitative training (ablation) | DiscoFaceGAN · FFHQ | PPL: 102 -> 123 | Hurt |
Closest papers by meaning
- Overcoming the Disentanglement vs Reconstruction Trade-off via Jacobian Supervision (ICLR 2019 · similarity 0.88)
- Disentangled Person Image Generation (CVPR 2018 · similarity 0.87)
- Learning Disentangled Representation by Exploiting Pretrained Generative Models: A Contrastive Learning View (ICLR 2022 · similarity 0.87)
- Explicit Disentanglement of Appearance and Perspective in Generative Models (NeurIPS 2019 · similarity 0.87)
- CapHuman: Capture Your Moments in Parallel Universes (CVPR 2024 · similarity 0.86)
- Transferring disentangled representations: bridging the gap between synthetic and real images (NeurIPS 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.