Reproduce “FabricGen: Microstructure-Aware Woven Fabric Generation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yingjie Tang, Di Luo, Zixiong Wang, Xiaoli Ling, jian Yang, Beibei Wang
FabricGen generates woven-fabric materials by separating broad color and pattern from fine weave structure, then combining them for rendering. A text-conditioned texture model produces the macro texture, while a language-model-guided procedural pipeline generates yarn-level detail. In comparisons with DressCode and MatFuse, its close-ups preserve finer structure, and it received 82.55% of user preferences; evidence is limited to tested woven-fabric cases.
Key insights
- Decomposing fabric material generation into macro-scale texture and micro-scale microstructure effectively addresses limitations of resolution and structural adherence in monolithic generative models.
- Large Language Models can be successfully fine-tuned and prompt-tuned with domain-specific expertise to automate complex structural design tasks, such as generating weaving drafts and fabric parameters from natural language.
- Accurately simulating multi-ply yarn configurations and subtle global irregularities (yarn sliding, flyaway fibers) through an enhanced procedural model is crucial for achieving photorealistic woven fabric appearance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.