Scribble-Guided Diffusion for Training-free Text-to-Image Generation

TL;DR

This paper introduces ScribbleDiff, a training-free method that uses simple scribbles as visual prompts to improve spatial control in text-to-image diffusion models, addressing limitations of existing guidance techniques.

Contribution

We propose a novel scribble-guided diffusion approach with moment alignment and scribble propagation, enabling precise spatial guidance without additional training.

Findings

Significant improvement in spatial control and alignment accuracy.

Effective guidance demonstrated on PASCAL-Scribble dataset.

Code availability facilitates reproducibility and further research.

Abstract

Recent advancements in text-to-image diffusion models have demonstrated remarkable success, yet they often struggle to fully capture the user's intent. Existing approaches using textual inputs combined with bounding boxes or region masks fall short in providing precise spatial guidance, often leading to misaligned or unintended object orientation. To address these limitations, we propose Scribble-Guided Diffusion (ScribbleDiff), a training-free approach that utilizes simple user-provided scribbles as visual prompts to guide image generation. However, incorporating scribbles into diffusion models presents challenges due to their sparse and thin nature, making it difficult to ensure accurate orientation alignment. To overcome these challenges, we introduce moment alignment and scribble propagation, which allow for more effective and flexible alignment between generated images and scribble inputs. Experimental results on the PASCAL-Scribble dataset demonstrate significant improvements in spatial control and consistency, showcasing the effectiveness of scribble-based guidance in diffusion models. Our code is available at https://github.com/kaist-cvml-lab/scribble-diffusion.