Research map / Label-efficient and robust learning / Domain adaptation and generalization
FreeDNA: Endowing Domain Adaptation of Diffusion-Based Dense Prediction with Training-Free Domain Noise Alignment
Hang Xu, Jie Huang, Linjiang Huang, Dong Li, Yidi Liu, Feng Zhao
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 DNA training-free alignment (source-available) | Marigold · RobotCar-night depth | AbsRel: 23.8 -> 17.0 | Improved |
| add DNA training-free alignment (source-available) | Lotus · RobotCar-night depth | delta1: 65.9 -> 78.1 | Improved |
| add source-free DNA | Marigold · NuScenes-night depth | AbsRel: 33.7 -> 28.8 | Improved |
| add DNA (source-available) | DDVM · FCDN optical flow | EPE: 7.41 -> 4.93 | Improved |
| add DNA (source-available) | DiffBIR · DRealSR blind SR | PSNR: 24.05 -> 25.19 | Improved |
| add source-free DNA | StableSR · DRealSR blind SR | MANIQA: 0.51 -> 0.51 | No clear effect |
| add DNA (source-available) | diffusion segmentation model · ACDC segmentation | mIOU: 53.7 -> 57.2 | Improved |
| use lambda_t scaling schedule instead of direct noise alignment (ablation) | DNA source-available · RobotCar-night depth | AbsRel: 25.5 -> 17.0 | Improved |
| use lambda_t scaling schedule instead of direct noise alignment (ablation) | DNA source-available · FCDN optical flow | EPE: 9.41 -> 4.93 | Improved |
| add linear mask schedule to alignment only (source-free) (ablation) | source-free DNA · RobotCar-night depth | AbsRel: 21.4 -> 18.5 | Improved |
| add noise consistency scaling to alignment only (source-free) (ablation) | source-free DNA · RobotCar-night depth | AbsRel: 21.4 -> 20.7 | Improved |
| add consistency scaling to linear mask schedule (ablation) | source-free DNA · RobotCar-night depth | AbsRel: 18.5 -> 18.2 | Improved |
Closest papers by meaning
- Vicinity-Guided Discriminative Latent Diffusion for Privacy-Preserving Domain Adaptation (NeurIPS 2025 · similarity 0.87)
- Everything to the Synthetic: Diffusion-driven Test-time Adaptation via Synthetic-Domain Alignment (CVPR 2025 · similarity 0.87)
- D-DOIT: Training-free Adaptation of Discrete Diffusion via Doob's h-Transform (NeurIPS 2026 · similarity 0.87)
- Domain Guidance: A Simple Transfer Approach for a Pre-trained Diffusion Model (ICLR 2025 · similarity 0.87)
- Diffusion-Driven Progressive Target Manipulation for Source-Free Domain Adaptation (NeurIPS 2025 · similarity 0.86)
- When Model Knowledge meets Diffusion Model: Diffusion-assisted Data-free Image Synthesis with Alignment of Domain and Class (ICML 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.
