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.
In un'altra lingua
dal contenuto originale
arxiv.org
Approfondimenti chiave tratti da
by Fangfu Liu,H... alle arxiv.org 03-15-2024
https://arxiv.org/pdf/2403.09625.pdfDomande più approfondite