So sánh phương pháp
Xem các phương pháp đã chọn cạnh nhau; những hàng khác biệt được làm nổi bật.
| Mô hình khuếch tán thích ứng miền× | Mô hình khuếch tán tinh chỉnh× | |
|---|---|---|
| Lĩnh vực | Học sâu | Học sâu |
| Họ | Machine learning | Machine learning |
| Năm ra đời≠ | 2022–2023 | 2020–2023 |
| Người khởi xướng≠ | Ho et al. (DDPM); domain-adaptation variants popularized by Gal et al. and Ruiz et al. (2022–2023) | Ho, J., Jain, A., Abbeel, P. (base DDPM); Ruiz et al. (DreamBooth fine-tuning paradigm) |
| Loại≠ | Generative model with domain adaptation | Generative model (fine-tuned via subject-specific or domain-specific data) |
| Công trình gốc≠ | Ho, J., Jain, A., & Abbeel, P. (2020). Denoising Diffusion Probabilistic Models. Advances in Neural Information Processing Systems, 33, 6840–6851. link ↗ | Ruiz, N., Li, Y., Jampani, V., Pritch, Y., Rubinstein, M., & Aberman, K. (2023). DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 22500–22510. DOI ↗ |
| Tên gọi khác | DA-diffusion model, domain-adapted diffusion model, domain-adaptive DDPM, cross-domain diffusion model | DDPM fine-tuning, diffusion model adaptation, personalized diffusion model, subject-driven diffusion fine-tuning |
| Liên quan≠ | 6 | 5 |
| Tóm tắt≠ | A domain-adaptive diffusion model is a denoising diffusion probabilistic model (DDPM) that is pre-trained on large general datasets and then adapted — through fine-tuning, textual inversion, or LoRA — to generate high-quality outputs in a specific target domain. It combines the powerful generative capacity of diffusion models with domain adaptation techniques, enabling high-fidelity synthesis in specialized areas such as medical imaging, satellite imagery, or domain-specific art styles with limited target-domain data. | A fine-tuned diffusion model adapts a large pretrained denoising diffusion model — such as Stable Diffusion or DALL-E — to a specific subject, style, or domain by continuing training on a small curated dataset. Techniques such as DreamBooth, textual inversion, and LoRA make this adaptation feasible on consumer hardware while preserving general generative capability. |
| ScholarGateBộ dữ liệu ↗ |
|
|