Research map / Label-efficient and robust learning / Domain adaptation and generalization
Few-Shot Image Generation via Cross-Domain Correspondence
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 cross-domain distance consistency during adaptation (ablation) | StyleGAN2 FFHQ adapted to caricatures · Caricatures | generation diversity | Helped |
| Use image-level adversarial loss for anchor-region samples and patch-level loss otherwise (ablation) | Few-shot image generation · Caricatures | image quality and diversity | Helped |
| Use image-level adversarial loss on all generations instead of the anchor-based image/patch setup (ablation) | Few-shot image generation with L_dist · Caricatures | part-level diversity | Hurt |
| Use patch-level adversarial loss on all generations instead of the anchor-based image/patch setup (ablation) | Few-shot image generation with L_dist · Caricatures | image quality | Hurt |
| Adapt with cross-domain distance consistency and relaxed realism | StyleGAN2 FFHQ · FFHQ-babies | FID: 87.41 $\pm$ 0.02 -> 74.39 $\pm$ 0.03 | Helped |
| Adapt with cross-domain distance consistency and relaxed realism | StyleGAN2 FFHQ · FFHQ-sunglasses | FID: 51.29 $\pm$ 0.05 -> 42.13 $\pm$ 0.04 | Helped |
| Adapt with cross-domain distance consistency and relaxed realism | StyleGAN2 FFHQ · Sketches | FID: 46.54 $\pm$ 0.01 -> 45.67 $\pm$ 0.02 | Helped |
| Adapt with cross-domain distance consistency and relaxed realism | StyleGAN2 FFHQ · Caricatures | intra-cluster pairwise LPIPS distance: 0.50 $\pm$ 0.05 -> 0.53 $\pm$ 0.01 | Helped |
| Adapt with cross-domain distance consistency and relaxed realism | StyleGAN2 FFHQ · Amedeo's paintings | intra-cluster pairwise LPIPS distance: 0.52 $\pm$ 0.03 -> 0.60 $\pm$ 0.01 | Helped |
| Adapt with cross-domain distance consistency and relaxed realism | StyleGAN2 FFHQ · Sketches | intra-cluster pairwise LPIPS distance: 0.42 $\pm$ 0.03 -> 0.45 $\pm$ 0.02 | Helped |
| Use cross-domain distance consistency and relaxed realism | StyleGAN2 FFHQ adapted to FFHQ-babies · FFHQ-babies | coverage: 0.325 -> 0.467 | Helped |
| Use cross-domain distance consistency and relaxed realism rather than MineGAN | StyleGAN2 FFHQ adapted to FFHQ-babies · FFHQ-babies | density: 0.803 -> 0.690 | Hurt |
Closest papers by meaning
- Few-shot Cross-domain Image Generation via Inference-time Latent-code Learning (ICLR 2023 · similarity 0.91)
- Few-shot Image Generation via Adaptation-Aware Kernel Modulation (NeurIPS 2022 · similarity 0.91)
- Towards Diverse and Faithful One-shot Adaption of Generative Adversarial Networks (NeurIPS 2022 · similarity 0.90)
- A Closer Look at Few-Shot Image Generation (CVPR 2022 · similarity 0.89)
- WeditGAN: Few-Shot Image Generation via Latent Space Relocation (AAAI 2024 · similarity 0.87)
- DomainGallery: Few-shot Domain-driven Image Generation by Attribute-centric Finetuning (NeurIPS 2024 · 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.