Research map / Label-efficient and robust learning / Domain adaptation and generalization
Transferring GANs: generating images from limited data
Yaxing Wang, Chenshen Wu, Luis Herranz, Joost van de Weijer, Abel Gonzalez-Garcia, Bogdan Raducanu
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 pre-trained discriminator to a scratch generator (ablation) | GAN · LSUN Bedrooms 100K | FID: 32.87 -> 30.57 | Improved |
| add pre-trained generator while discriminator stays scratch (ablation) | GAN · LSUN Bedrooms 100K | FID: 32.87 -> 56.16 | Hurt |
| transfer both generator and discriminator (ablation) | GAN · LSUN Bedrooms 100K | FID: 32.87 -> 24.35 | Improved |
| transfer both networks, measured by IW (ablation) | GAN · LSUN Bedrooms 100K | IW: -4.27 -> -3.88 | Improved |
| pre-train on ImageNet with only 1K target images | GAN · LSUN Bedrooms 1K | FID: 256.1 -> 93.4 | Improved |
| pre-train on ImageNet with 1M target images | GAN · LSUN Bedrooms 1M | FID: 21.0 -> 18.5 | Improved |
| pre-train with 100K target images | GAN · LSUN Bedrooms 100K | FID: 32.9 -> 24.4 | Improved |
| use CelebA as source instead of scratch | GAN · LFW | FID: 19.36 -> 7.16 | Improved |
| use ImageNet as source instead of scratch | GAN · Flowers | FID: 71.98 -> 54.04 | Improved |
| use Bedrooms as source for Kitchens instead of Places | GAN · Kitchens | FID: 34.59 -> 28.54 | Improved |
| transfer from various sources for Cityscapes | GAN · Cityscapes | FID: 155.68 -> 122.46 | No clear effect |
| initialize AC-GAN from Places GAN, 250 iterations | AC-GAN · LSUN 10 classes | FID: 352.4 -> 107.2 | Improved |
Closest papers by meaning
- One-Shot Generative Domain Adaptation (ICCV 2023 · similarity 0.94)
- Unsupervised Pixel-Level Domain Adaptation With Generative Adversarial Networks (CVPR 2017 · similarity 0.92)
- On Leveraging Pretrained GANs for Generation with Limited Data (ICML 2020 · similarity 0.92)
- Generalized One-shot Domain Adaptation of Generative Adversarial Networks (NeurIPS 2022 · similarity 0.92)
- Efficient Conditional GAN Transfer With Knowledge Propagation Across Classes (CVPR 2021 · similarity 0.91)
- Generate to Adapt: Aligning Domains Using Generative Adversarial Networks (CVPR 2018 · similarity 0.91)
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.
