Research map / Image generation / Autoregressive and latent generation
Where Is My Spot? Few-Shot Image Generation via Latent Subspace Optimization
Chenxi Zheng, Bangzhen Liu, Huaidong Zhang, Xuemiao Xu, Shengfeng He
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 latent subspace refinement (LSR) to the framework (ablation) | latent subspace optimization · Flowers, 3-shot | FID: 37.94 -> 34.59 (w/o LSR vs full) | Improved |
| add latent anchor localization (LAL) (ablation) | latent subspace optimization · Flowers, 3-shot | FID: 36.22 -> 34.59 (w/o LAL vs full) | Improved |
| add latent anchor localization (LAL) (ablation) | latent subspace optimization · Flowers, 3-shot | LPIPS (diversity): 0.3884 (w/o LAL) -> 0.3914 full is not lower; LAL limits diversity per text | Hurt |
| add regularization with seen semantics (ablation) | latent subspace optimization · Flowers, 3-shot | FID: 53.13 -> 34.59 (w/o Reg vs full) | Improved |
| add latent subspace refinement (LSR) (ablation) | latent subspace optimization · Flowers, 3-shot | time cost (s): 11 s (w/o LSR) -> 136 s full | Hurt |
| latent subspace optimization vs. prior methods (one-shot) | conditional StyleGAN · Flowers / AnimalFaces / VGGFaces, 1-shot | FID: FID reduced by 7.58 (Flowers) and 4.37 (AnimalFaces) vs. best baseline; 0.98 on VGGFaces at 3-shot | Improved |
| latent subspace optimization vs. prior methods (3-shot) | conditional StyleGAN · Flowers / AnimalFaces / VGGFaces, 3-shot | LPIPS (diversity): Flowers 0.3914 vs WaveGAN 0.3868; VGGFaces 0.3344 vs LoFGAN 0.2869 but below AGE/WaveGAN 1-shot | Mixed |
| add perceptual/semantic stabilization loss on anchors | latent subspace refinement · qualitative | semantic preservation | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- LoFGAN — LoFGAN: Fusing Local Representations for Few-Shot Image Generation (ICCV 2021) · better on 2 of 2 shared results
- DAGAN — DA-GAN: Instance-Level Image Translation by Deep Attention Generative Adversarial Networks (CVPR 2018) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Low-Rank Subspaces in GANs (NeurIPS 2021 · similarity 0.87)
- WeditGAN: Few-Shot Image Generation via Latent Space Relocation (AAAI 2024 · similarity 0.86)
- Plug & Play Generative Networks: Conditional Iterative Generation of Images in Latent Space (CVPR 2017 · similarity 0.86)
- Diversity-aware Channel Pruning for StyleGAN Compression (CVPR 2024 · similarity 0.86)
- Linear combinations of latents in generative models: subspaces and beyond (ICLR 2025 · similarity 0.85)
- Stabilize the Latent Space for Image Autoregressive Modeling: A Unified Perspective (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.
