SDiT: Semantic Region-Adaptive for Diffusion Transformers
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918294774087680 |
|---|---|
| author | Lin, Bowen Ye, Fanjiang Liu, Yihua Guo, Zhenghui Zhang, Boyuan Zheng, Weijian Xu, Yufan Xing, Tiancheng Wang, Yuke Zhang, Chengming |
| author_facet | Lin, Bowen Ye, Fanjiang Liu, Yihua Guo, Zhenghui Zhang, Boyuan Zheng, Weijian Xu, Yufan Xing, Tiancheng Wang, Yuke Zhang, Chengming |
| contents | Diffusion Transformers (DiTs) achieve state-of-the-art performance in text-to-image synthesis but remain computationally expensive due to the iterative nature of denoising and the quadratic cost of global attention. In this work, we observe that denoising dynamics are spatially non-uniform-background regions converge rapidly while edges and textured areas evolve much more actively. Building on this insight, we propose SDiT, a Semantic Region-Adaptive Diffusion Transformer that allocates computation according to regional complexity. SDiT introduces a training-free framework combining (1) semantic-aware clustering via fast Quickshift-based segmentation, (2) complexity-driven regional scheduling to selectively update informative areas, and (3) boundary-aware refinement to maintain spatial coherence. Without any model retraining or architectural modification, SDiT achieves up to 3.0x acceleration while preserving nearly identical perceptual and semantic quality to full-attention inference. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_12283 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | SDiT: Semantic Region-Adaptive for Diffusion Transformers Lin, Bowen Ye, Fanjiang Liu, Yihua Guo, Zhenghui Zhang, Boyuan Zheng, Weijian Xu, Yufan Xing, Tiancheng Wang, Yuke Zhang, Chengming Computer Vision and Pattern Recognition Diffusion Transformers (DiTs) achieve state-of-the-art performance in text-to-image synthesis but remain computationally expensive due to the iterative nature of denoising and the quadratic cost of global attention. In this work, we observe that denoising dynamics are spatially non-uniform-background regions converge rapidly while edges and textured areas evolve much more actively. Building on this insight, we propose SDiT, a Semantic Region-Adaptive Diffusion Transformer that allocates computation according to regional complexity. SDiT introduces a training-free framework combining (1) semantic-aware clustering via fast Quickshift-based segmentation, (2) complexity-driven regional scheduling to selectively update informative areas, and (3) boundary-aware refinement to maintain spatial coherence. Without any model retraining or architectural modification, SDiT achieves up to 3.0x acceleration while preserving nearly identical perceptual and semantic quality to full-attention inference. |
| title | SDiT: Semantic Region-Adaptive for Diffusion Transformers |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2601.12283 |