Research map / Image generation / GANs
Few-Shot Image Generation with Mixup-Based Distance Learning
Chaerin Kong, Jeesoo Kim, Donghoon Han, Nojun Kwak
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 MixDL (latent mixup and distance learning) | StyleGAN2 · Anime Face 10-shot | FID: 166.0 -> 73.1 | Improved |
| add MixDL | StyleGAN2 · Pokemon 10-shot | FID: 257.6 -> 117.0 | Improved |
| add MixDL | FastGAN · Anime Face 10-shot | FID: 123.7 -> 107.6 | Improved |
| add MixDL | FastGAN · Pokemon 10-shot | FID: 123.5 -> 144.0 | Hurt |
| add MixDL | FastGAN · Oxford Flowers 10-shot | FID: 182.7 -> 180.5 | No clear effect |
| add MixDL on top of ADA | StyleGAN2 + ADA · Animal-Face Dog 10-shot | FID: 236.5 -> 94.1 | Improved |
| add MixDL on top of DA | StyleGAN2 + DA · Face Sketches 10-shot | FID: 43.1 -> 35.6 | Improved |
| replace DistanceGAN with MixDL | StyleGAN2 · FFHQ-babies 10-shot | FID: 105.7 -> 83.4 | Improved |
| use MixDL instead of CDC pretrained on FFHQ | StyleGAN2 · Obama 10-shot | FID: 75.0 -> 62.7 | Improved |
| add MixDL to StyleGAN2 | StyleGAN2 · Landscape interpolation | PPL std: 12.66 -> 4.19 | Improved |
| use 8 FC mapping layers instead of 2 in DiffAug | DiffAug · Obama 100-shot | FID: 46.87 -> 71.13 | Hurt |
| add discriminator feature alignment (MixDL-D) to MixDL-G (ablation) | MixDL · few-shot datasets | FID/diversity | Mixed |
Closest papers by meaning
- Few-Shot Image Generation via Cross-Domain Correspondence (CVPR 2021 · similarity 0.89)
- 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 0.88)
- Smoothness Similarity Regularization for Few-Shot GAN Adaptation (ICCV 2023 · similarity 0.88)
- MS$^3$D: A RG Flow-Based Regularization for GAN Training with Limited Data (ICML 2024 · similarity 0.86)
- Diverse Rare Sample Generation with Pretrained GANs (AAAI 2025 · 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.
