Customize Your Own Paired Data via Few-shot Way
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866914803858014208 |
|---|---|
| author | Chen, Jinshu Li, Bingchuan Hua, Miao Xu, Panpan He, Qian |
| author_facet | Chen, Jinshu Li, Bingchuan Hua, Miao Xu, Panpan He, Qian |
| contents | Existing solutions to image editing tasks suffer from several issues. Though achieving remarkably satisfying generated results, some supervised methods require huge amounts of paired training data, which greatly limits their usages. The other unsupervised methods take full advantage of large-scale pre-trained priors, thus being strictly restricted to the domains where the priors are trained on and behaving badly in out-of-distribution cases. The task we focus on is how to enable the users to customize their desired effects through only few image pairs. In our proposed framework, a novel few-shot learning mechanism based on the directional transformations among samples is introduced and expands the learnable space exponentially. Adopting a diffusion model pipeline, we redesign the condition calculating modules in our model and apply several technical improvements. Experimental results demonstrate the capabilities of our method in various cases. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2405_12490 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Customize Your Own Paired Data via Few-shot Way Chen, Jinshu Li, Bingchuan Hua, Miao Xu, Panpan He, Qian Computer Vision and Pattern Recognition Existing solutions to image editing tasks suffer from several issues. Though achieving remarkably satisfying generated results, some supervised methods require huge amounts of paired training data, which greatly limits their usages. The other unsupervised methods take full advantage of large-scale pre-trained priors, thus being strictly restricted to the domains where the priors are trained on and behaving badly in out-of-distribution cases. The task we focus on is how to enable the users to customize their desired effects through only few image pairs. In our proposed framework, a novel few-shot learning mechanism based on the directional transformations among samples is introduced and expands the learnable space exponentially. Adopting a diffusion model pipeline, we redesign the condition calculating modules in our model and apply several technical improvements. Experimental results demonstrate the capabilities of our method in various cases. |
| title | Customize Your Own Paired Data via Few-shot Way |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2405.12490 |