Research map / Label-efficient and robust learning / Domain adaptation and generalization
Text-Driven Generative Domain Adaptation with Spectral Consistency Regularization
Zhenhuan Liu, Liang Li, Jiayu Xiao, Zheng-Jun Zha, Qingming Huang
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 spectral consistency regularization to directional-loss adaptation | StyleGAN-NADA · Photo->Sketch | Hessian trace: 0.061 -> 0.116 | Improved |
| add spectral consistency regularization to directional-loss adaptation | StyleGAN-NADA · Photo->Sketch | PPL: 323.25 -> 463.57 | Improved |
| add spectral consistency regularization to directional-loss adaptation | StyleGAN-NADA · Human->Werewolf | PPL: 302.91 -> 383.19 | Improved |
| use SCC, WDC or MS regularization instead of spectral consistency | text-driven GAN adaptation · diversity vs stylization | stylization (qualitative) | Hurt |
| use PPL or CDC regularization instead of spectral consistency | text-driven GAN adaptation · diversity | Hessian trace / PPL: Photo->Sketch trace 0.028 (PPL reg), 0.062 (CDC) vs 0.116 (Ours) | Hurt |
| apply regularization only to coarse, only to fine, or global granularity instead of adaptive soft mask (ablation) | Granularity adaptive regularization · Photo->Edvard Munch (qualitative) | diversity and style | Improved |
| increase lambda_spectral (ablation) | Spectral consistency regularization · qualitative samples | diversity | Improved |
| use CLIP image encoder as metric space instead of VGG, MoCo, ArcFace or pixel space (ablation) | Spectral consistency regularization · identity preservation (qualitative) | identity of source image | Improved |
| add spectral consistency regularization versus WDC in user study | Ours vs WDC · user study, attribute consistency | preference (%): 49.8% | No clear effect |
Closest papers by meaning
- Navigating Text-To-Image Customization: From LyCORIS Fine-Tuning to Model Evaluation (ICLR 2024 · similarity 0.86)
- Few-shot Hybrid Domain Adaptation of Image Generator (ICLR 2024 · similarity 0.85)
- Moment- and Power-Spectrum-Based Gaussianity Regularization for Text-to-Image Models (NeurIPS 2025 · similarity 0.85)
- Progressive Text-to-Image Diffusion with Soft Latent Direction (AAAI 2024 · similarity 0.85)
- Direct Consistency Optimization for Robust Customization of Text-to-Image Diffusion models (NeurIPS 2024 · similarity 0.85)
- Invertible Projection and Conditional Alignment for Multi-Source Blended-Target Domain Adaptation (AAAI 2025 · 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.
