The authors developed a sparse latent flow-matching model that generates 3D shapes represented as sparse voxel grids, operating at an effective resolution of 2048³.
Sparse flow matching generates 3D shapes at ultra-high voxel resolution
Filigree3D produces 2048³ geometry from a single image in about one minute, with straightforward extension to 4096³.
Chinese Tech
Hongjie Li · Xinran Yang · Xiuchao Wu · Jiangjing Lyu · Chengfei Lv
Alibaba Group
Research Digest··2 min read
The authors introduce Filigree3D, a sparse latent flow-matching framework that generates 3D geometry from a single image at voxel resolutions up to 2048³, scalable to 4096³.
Why this paper
From Alibaba Group
In one line
Filigree3D generates detailed 3D geometry from one image at 2048^3 voxel resolution by combining sparse computation, local-global attention, and multi-scale visual conditioning.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§