Controllable Segmentation-Based Text-Guided Style Editing
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866912509106061312 |
|---|---|
| author | Li, Jingwen Chandrasekar, Aravind Rocha, Mariana Li, Chao Chen, Yuqing |
| author_facet | Li, Jingwen Chandrasekar, Aravind Rocha, Mariana Li, Chao Chen, Yuqing |
| contents | We present a novel approach for controllable, region-specific style editing driven by textual prompts. Building upon the state-space style alignment framework introduced by \emph{StyleMamba}, our method integrates a semantic segmentation model into the style transfer pipeline. This allows users to selectively apply text-driven style changes to specific segments (e.g., ``turn the building into a cyberpunk tower'') while leaving other regions (e.g., ``people'' or ``trees'') unchanged. By incorporating region-wise condition vectors and a region-specific directional loss, our method achieves high-fidelity transformations that respect both semantic boundaries and user-driven style descriptions. Extensive experiments demonstrate that our approach can flexibly handle complex scene stylizations in real-world scenarios, improving control and quality over purely global style transfer methods. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_16129 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Controllable Segmentation-Based Text-Guided Style Editing Li, Jingwen Chandrasekar, Aravind Rocha, Mariana Li, Chao Chen, Yuqing Graphics We present a novel approach for controllable, region-specific style editing driven by textual prompts. Building upon the state-space style alignment framework introduced by \emph{StyleMamba}, our method integrates a semantic segmentation model into the style transfer pipeline. This allows users to selectively apply text-driven style changes to specific segments (e.g., ``turn the building into a cyberpunk tower'') while leaving other regions (e.g., ``people'' or ``trees'') unchanged. By incorporating region-wise condition vectors and a region-specific directional loss, our method achieves high-fidelity transformations that respect both semantic boundaries and user-driven style descriptions. Extensive experiments demonstrate that our approach can flexibly handle complex scene stylizations in real-world scenarios, improving control and quality over purely global style transfer methods. |
| title | Controllable Segmentation-Based Text-Guided Style Editing |
| topic | Graphics |
| url | https://arxiv.org/abs/2503.16129 |