DuoLoRA : Cycle-consistent and Rank-disentangled Content-Style Personalization
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908325639094272 |
|---|---|
| author | Roy, Aniket Borse, Shubhankar Kadambi, Shreya Das, Debasmit Mahajan, Shweta Garrepalli, Risheek Park, Hyojin Nayak, Ankita Chellappa, Rama Hayat, Munawar Porikli, Fatih |
| author_facet | Roy, Aniket Borse, Shubhankar Kadambi, Shreya Das, Debasmit Mahajan, Shweta Garrepalli, Risheek Park, Hyojin Nayak, Ankita Chellappa, Rama Hayat, Munawar Porikli, Fatih |
| contents | We tackle the challenge of jointly personalizing content and style from a few examples. A promising approach is to train separate Low-Rank Adapters (LoRA) and merge them effectively, preserving both content and style. Existing methods, such as ZipLoRA, treat content and style as independent entities, merging them by learning masks in LoRA's output dimensions. However, content and style are intertwined, not independent. To address this, we propose DuoLoRA, a content-style personalization framework featuring three key components: (i) rank-dimension mask learning, (ii) effective merging via layer priors, and (iii) Constyle loss, which leverages cycle-consistency in the merging process. First, we introduce ZipRank, which performs content-style merging within the rank dimension, offering adaptive rank flexibility and significantly reducing the number of learnable parameters. Additionally, we incorporate SDXL layer priors to apply implicit rank constraints informed by each layer's content-style bias and adaptive merger initialization, enhancing the integration of content and style. To further refine the merging process, we introduce Constyle loss, which leverages the cycle-consistency between content and style. Our experimental results demonstrate that DuoLoRA outperforms state-of-the-art content-style merging methods across multiple benchmarks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_13206 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | DuoLoRA : Cycle-consistent and Rank-disentangled Content-Style Personalization Roy, Aniket Borse, Shubhankar Kadambi, Shreya Das, Debasmit Mahajan, Shweta Garrepalli, Risheek Park, Hyojin Nayak, Ankita Chellappa, Rama Hayat, Munawar Porikli, Fatih Graphics We tackle the challenge of jointly personalizing content and style from a few examples. A promising approach is to train separate Low-Rank Adapters (LoRA) and merge them effectively, preserving both content and style. Existing methods, such as ZipLoRA, treat content and style as independent entities, merging them by learning masks in LoRA's output dimensions. However, content and style are intertwined, not independent. To address this, we propose DuoLoRA, a content-style personalization framework featuring three key components: (i) rank-dimension mask learning, (ii) effective merging via layer priors, and (iii) Constyle loss, which leverages cycle-consistency in the merging process. First, we introduce ZipRank, which performs content-style merging within the rank dimension, offering adaptive rank flexibility and significantly reducing the number of learnable parameters. Additionally, we incorporate SDXL layer priors to apply implicit rank constraints informed by each layer's content-style bias and adaptive merger initialization, enhancing the integration of content and style. To further refine the merging process, we introduce Constyle loss, which leverages the cycle-consistency between content and style. Our experimental results demonstrate that DuoLoRA outperforms state-of-the-art content-style merging methods across multiple benchmarks. |
| title | DuoLoRA : Cycle-consistent and Rank-disentangled Content-Style Personalization |
| topic | Graphics |
| url | https://arxiv.org/abs/2504.13206 |