Filigree3D generates ultra-high-resolution 3D models from images fast
FILIGREE3D: Scaling Sparse Latent Flow Matching for Ultra-High-Resolution Image-to-3D Generation
Computer Vision and Pattern Recognition
Summary
Creating very detailed 3D models from just one image is usually slow and uses a lot of computer memory. The authors designed Filigree3D, a method that can make extremely detailed 3D models with voxel sizes up to 2048 cubed, running efficiently on current GPUs. They use clever ways to focus processing only where details matter and combine structural information with image features to improve fine details. This approach helps create complex 3D shapes quickly without losing important details compared to older methods.
What this means in practice
- •For 3d artists and modelers: Create ultra-detailed 3D models from single images efficiently for high-quality visual effects and design projects.
- •For game developers: Generate detailed 3D assets from images fast while keeping memory use manageable for game environments.
Authors
Hongjie Li, Xinran Yang, Xiuchao Wu, Jiangjing Lyu, Chengfei Lv
Abstract
Scaling image-to-3D generation to ultra-high resolutions requires controlling rapidly growing computational costs without sacrificing fine geometric detail. We present \textbf{Filigree3D}, a sparse latent flow-matching framework that generates 3D geometry from a single image at voxel resolutions up to $2048^3$, with straightforward extensibility to $4096^3$. To make training tractable, we introduce Structure-Aware Sparse Scaling, which combines spatial bounding with alternating local-global attention to constrain token growth while preserving both fine-scale details and long-range structural context. To enhance detail reconstruction, we curate training samples based on their high-resolution geometric gains and inject multi-scale image features into a sparse 3D DiT, effectively coupling structural semantics with fine-grained visual cues. Furthermore, a visibility-aware voxel regularization strategy improves robustness against sparse perturbations and facilitates the completion of unobserved geometry. Under our default configuration, Filigree3D maintains peak GPU memory consumption within practical limits for contemporary hardware, enabling the generation of highly intricate 3D geometry in approximately one minute. Extensive experiments demonstrate that our method yields substantial improvements in overall geometric fidelity and fine-detail preservation compared to existing baselines, validating practical, detail-preserving 3D generation at unprecedented resolutions.