Research map / Image generation / GANs
LFS-GAN: Lifelong Few-Shot Image Generation
Juwon Seo, Ji-Su Kang, Gyeong-Moon Park
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 LeFT modulation with frozen weights instead of finetuning (lifelong setting) | StyleGAN2 · Lifelong few-shot generation | average FID: 98.66 -> 33.77 | Improved |
| add bias term to LeFT (rank 1) (ablation) | LeFT · Lifelong few-shot generation, average | FID: 39.42 -> 33.77 | Improved |
| use rank 1 instead of higher ranks (ablation) | LeFT · Lifelong few-shot generation, average | FID: 55.20 -> 33.77 (r=16 to r=1) | Improved |
| use ReLU activation in LeFT instead of identity (ablation) | LeFT · Lifelong few-shot generation, average | FID: 39.87 -> 33.77 | Improved |
| use ReLU instead of SiLU (ablation) | LeFT · Lifelong few-shot generation, average | FID: 40.28 -> 33.77 | Improved |
| add cluster-wise mode seeking loss (ablation) | LFS-GAN · Lifelong few-shot generation, average | B-LPIPS: 0.423 -> 0.489 | Improved |
| maximize only a single term of the mode seeking loss instead of all three (ablation) | LFS-GAN · Lifelong few-shot generation, average | B-LPIPS: 0.436 -> 0.489 | Improved |
| apply LFS-GAN to the independent few-shot generation task | LFS-GAN · Few-shot generation (FFHQ source) | average FID: 44.62 -> 33.77 | Improved |
| use LeFT low-rank factorization to cut trainable parameters | LFS-GAN · Parameter efficiency | trainable parameters per task: 30.0M -> 0.1M | Improved |
Closest papers by meaning
- Lifelong GAN: Continual Learning for Conditional Image Generation (ICCV 2019 · similarity 0.89)
- Piggyback GAN: Efficient Lifelong Learning for Image Conditioned Generation (ECCV 2020 · similarity 0.89)
- A Closer Look at Few-Shot Image Generation (CVPR 2022 · similarity 0.89)
- LoFGAN: Fusing Local Representations for Few-Shot Image Generation (ICCV 2021 · similarity 0.89)
- Few-shot Image Generation via Adaptation-Aware Kernel Modulation (NeurIPS 2022 · similarity 0.88)
- WeditGAN: Few-Shot Image Generation via Latent Space Relocation (AAAI 2024 · similarity 0.87)
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.
