SketchDirector enables complex scene sketching
Original title: SketchDirector: Complex Scene Sketching via Agentic Layout Reasoning and Detail Refinement
SketchDirector uses LLM decomposition and diffusion refinement to mimic human sketching workflow, introducing MCSketch8K dataset.
Existing sketch generation methods rely on image guidance and are limited to individual objects or simplified scenes. SketchDirector employs a large language model to decompose scene-level text prompts into multiple draft assets represented by Bézier curves. An agentic reasoning loop iteratively plans their spatial arrangement guided by self-evolving memory, followed by a detail refinement module that leverages diffusion priors to enhance sketch quality while preserving geometric structure. Extensive experiments show that SketchDirector outperforms state-of-the-art baselines in semantic and compositional quality.
- Organizations
- Hugging Face
- AI models
- SketchDirector
From the original article
1 min read at huggingface.co
Existing sketch generation methods often rely on image guidance and are largely limited to individual objects or simplified scenes, leaving text-guided complex scene sketching underexplored. We present SketchDirector, an agentic framework that mimics the human sketching workflow for automatic multi-concept scene sketching.
By Yiqi Liang, Xiaodong Cun · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.