Research map / 3D vision / Face and shape
SparseFlex: High-Resolution and Arbitrary-Topology 3D Shape Modeling
Xianglong He, Zi-Xin Zou, Chia-Hao Chen, Yuan-Chen Guo, Ding Liang, Chun Yuan, Wanli Ouyang, Yan-Pei Cao, Yangguang Li
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| increase VAE resolution from 256 to 1024 (ablation) | SparseFlex VAE · Toys4K (entire dataset) | Chamfer distance: 2.56 -> 1.33 | Improved |
| add self-pruning upsampling module (ablation) | SparseFlex VAE (256) · Deepfashion3D | Chamfer distance: 0.55 -> 0.08 | Improved |
| add self-pruning upsampling module (ablation) | SparseFlex VAE (1024) · Deepfashion3D | F1(0.001): 24.80 -> 37.22 | Improved |
| add frustum-aware sectional voxel training (ablation) | SparseFlex · training at 1024^3 | GPU memory / OOM: OOM -> 55441 MB | Improved |
| add sparse structure to Flexicubes without frustum training (ablation) | SparseFlex · training at 512^3 and 1024^3 | GPU memory: OOM at 512^3 -> 62029 MB | Improved |
| lower visibility ratio alpha from 0.3 to 0.1 (ablation) | SparseFlex training · 1024^3 | GPU memory (MB): 69991 -> 55441 | Improved |
| use SparseFlex VAE for structured latent flow instead of TRELLIS | image-to-3D generation · Toys4k | FID: 47.66 -> 44.95 | Improved |
| compare to 3PSDF on open surface | SparseFlex 256 vs 3PSDF · Deepfashion3D | Chamfer distance: 3PSDF 0.26 vs Ours 256 0.08; 3PSDF F1(0.01) 99.35 vs 99.99 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- 3PSDF* — 3PSDF: Three-Pole Signed Distance Function for Learning Surfaces With Arbitrary Topologies (CVPR 2022) · better on 4 of 4 shared results
- XCube — XCube: Large-Scale 3D Generative Modeling using Sparse Voxel Hierarchies (CVPR 2024) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Sparc3D: Sparse Representation and Construction for High-Resolution 3D Shapes Modeling (NeurIPS 2025 · similarity 0.91)
- MeshFormer : High-Quality Mesh Generation with 3D-Guided Reconstruction Model (NeurIPS 2024 · similarity 0.90)
- GeoSVR: Taming Sparse Voxels for Geometrically Accurate Surface Reconstruction (NeurIPS 2025 · similarity 0.88)
- SILSA: Sliding-Window Slice Latents for Topology-Preserving High-Resolution 3D Generation (NeurIPS 2026 · similarity 0.87)
- Direct3D-S2: Gigascale 3D Generation Made Easy with Spatial Sparse Attention (NeurIPS 2025 · similarity 0.87)
- Ghost on the Shell: An Expressive Representation of General 3D Shapes (ICLR 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
