The paper introduces Make-Your-3D, a method for personalized 3D content creation. It addresses the limitations of existing methods in generating subject-driven 3D content. The key insight is to harmonize the distributions of a multi-view diffusion model and an identity-specific 2D generative model. The co-evolution framework reduces distribution variance through identity-aware optimization and subject-prior optimization. Extensive experiments demonstrate high-quality, consistent, and subject-specific 3D content generation within minutes.
Recent advancements in 3D generation models have led to new creative possibilities by allowing users to guide the process through a single image or natural language.
Quantitative comparisons with DreamBooth3D and MV DreamBooth show higher CLIP R-Precision scores for Make-Your-3D. User study results indicate preference for Make-Your-3D over other methods in terms of multi-view consistency, subject fidelity, prompt fidelity, and overall quality.
Make-Your-3D offers efficient and effective personalized 3D content creation, demonstrating potential applications across various sectors.
다른 언어로
소스 콘텐츠 기반
arxiv.org
더 깊은 질문