Stroke2Sketch: Harnessing Stroke Attributes for Training-Free Sketch Generation
Rui Yang, Huining Li, Yiyi Long, Xiaojun Wu, Shengfeng He · 2025
Read Stroke2Sketch: Harnessing Stroke Attributes for Training-Free Sketch Generation by Rui Yang, Huining Li, Yiyi Long, Xiaojun Wu, Shengfeng He (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-16319.txt.
Description
Rui Yang and four co-authors generate sketches that copy not just a picture's content but the hand of a reference artist. Faithful style transfer for drawing means carrying over stroke qualities like line thickness, wobble, and texture sparsity while keeping the subject intact, a finer demand than most style methods handle.
Stroke2Sketch needs no training. Its cross-image stroke attention, inserted into the self-attention layers of a generative model, builds fine-grained correspondences between content and reference so stroke attributes land in the right places. Adaptive contrast enhancement and semantic-focused attention reinforce the subject and foreground. The authors report sketches that look handcrafted, beating earlier methods on expressive stroke control and semantic coherence, and they release the code publicly.
Catalog record
| CREATOR | Rui Yang, Huining Li, Yiyi Long, Xiaojun Wu, Shengfeng He |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-16319