FocalDreamer: Text-driven 3D Editing via Focal-fusion Assembly

Abstract: While text-3D editing has made significant strides in leveraging score distillation sampling, emerging approaches still fall short in delivering separable, precise and consistent outcomes that are vital to content creation. In response, we introduce FocalDreamer, a framework that merges base shape with editable parts according to text prompts for fine-grained editing within desired regions. Specifically, equipped with geometry union and dual-path rendering, FocalDreamer assembles independent 3D parts into a complete object, tailored for convenient instance reuse and part-wise control. We propose geometric focal loss and style consistency regularization, which encourage focal fusion and congruent overall appearance. Furthermore, FocalDreamer generates high-fidelity geometry and PBR textures which are compatible with widely-used graphics engines. Extensive experiments have highlighted the superior editing capabilities of FocalDreamer in both quantitative and qualitative evaluations.

• The paper introduces a novel text-driven 3D editing framework that merges base shapes with editable parts via focal fusion.
• The methodology leverages geometric focal loss and style consistency regularization to ensure seamless integration and uniform appearance.
• Extensive experiments demonstrate high-fidelity geometry, PBR textures, and precise part-wise control for versatile content creation.

"FocalDreamer: Text-driven 3D Editing via Focal-fusion Assembly" addresses the challenges in the field of text-driven 3D editing, particularly focusing on achieving separable, precise, and consistent outcomes necessary for content creation. Traditional methods that employ score distillation sampling have struggled to deliver these required attributes. FocalDreamer introduces a novel framework that adeptly combines base shapes with editable components based on text prompts, enabling fine-grained editing focused on specific regions.

The framework utilizes a combination of geometry union and dual-path rendering to amalgamate independent 3D parts into a cohesive whole. This approach is designed to facilitate instance reuse and allow for detailed part-wise control, which is crucial for efficient and versatile content creation workflows.

Central to FocalDreamer's innovation are two key concepts: geometric focal loss and style consistency regularization. The geometric focal loss aims to enhance focal fusion, ensuring that the merged parts integrate seamlessly. Meanwhile, style consistency regularization works to maintain a congruent overall appearance, ensuring that the edited regions are stylistically consistent with the rest of the object.

An important aspect of FocalDreamer is its ability to generate high-fidelity geometry and physically-based rendering (PBR) textures. This compatibility with widely-used graphics engines ensures that the generated content can be easily integrated into existing pipelines, making the framework both practical and broadly applicable.

The paper reports extensive experiments that demonstrate FocalDreamer's superior editing capabilities. These experiments include both quantitative metrics and qualitative evaluations, highlighting the framework's ability to deliver high-quality, precise, and consistent 3D content based on textual descriptions. As such, FocalDreamer represents a significant advancement in the field of text-driven 3D content creation, providing a robust tool for artists and designers to achieve detailed and coherent 3D edits.