DREAM: Diffusion Rectification and Estimation-Adaptive Models
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2023
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866909142857285632 |
|---|---|
| author | Zhou, Jinxin Ding, Tianyu Chen, Tianyi Jiang, Jiachen Zharkov, Ilya Zhu, Zhihui Liang, Luming |
| author_facet | Zhou, Jinxin Ding, Tianyu Chen, Tianyi Jiang, Jiachen Zharkov, Ilya Zhu, Zhihui Liang, Luming |
| contents | We present DREAM, a novel training framework representing Diffusion Rectification and Estimation Adaptive Models, requiring minimal code changes (just three lines) yet significantly enhancing the alignment of training with sampling in diffusion models. DREAM features two components: diffusion rectification, which adjusts training to reflect the sampling process, and estimation adaptation, which balances perception against distortion. When applied to image super-resolution (SR), DREAM adeptly navigates the tradeoff between minimizing distortion and preserving high image quality. Experiments demonstrate DREAM's superiority over standard diffusion-based SR methods, showing a $2$ to $3\times $ faster training convergence and a $10$ to $20\times$ reduction in sampling steps to achieve comparable results. We hope DREAM will inspire a rethinking of diffusion model training paradigms. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2312_00210 |
| institution | arXiv |
| publishDate | 2023 |
| record_format | arxiv |
| spellingShingle | DREAM: Diffusion Rectification and Estimation-Adaptive Models Zhou, Jinxin Ding, Tianyu Chen, Tianyi Jiang, Jiachen Zharkov, Ilya Zhu, Zhihui Liang, Luming Computer Vision and Pattern Recognition Artificial Intelligence We present DREAM, a novel training framework representing Diffusion Rectification and Estimation Adaptive Models, requiring minimal code changes (just three lines) yet significantly enhancing the alignment of training with sampling in diffusion models. DREAM features two components: diffusion rectification, which adjusts training to reflect the sampling process, and estimation adaptation, which balances perception against distortion. When applied to image super-resolution (SR), DREAM adeptly navigates the tradeoff between minimizing distortion and preserving high image quality. Experiments demonstrate DREAM's superiority over standard diffusion-based SR methods, showing a $2$ to $3\times $ faster training convergence and a $10$ to $20\times$ reduction in sampling steps to achieve comparable results. We hope DREAM will inspire a rethinking of diffusion model training paradigms. |
| title | DREAM: Diffusion Rectification and Estimation-Adaptive Models |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2312.00210 |