Research map / Point clouds / Shape and surface completion
HilbertGen-3D:Hilbert--Multifractal Conditioning for Topology-Aware 3D Generation
Ziyi Zhang, Shixuan Li, Yorie Nakahira, Paul Bogdan, Guannan Qu
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes HilbertGen-3D, a diffusion-based 3D point cloud generator conditioned on Hilbert-curve and multifractal descriptors to capture global topological structure.
Problem
Diffusion-based point cloud generation relies on local voxel or patch features and often misses global structural organization, which matters for game assets, medical imaging, and VR/AR.
Approach
Conditions a diffusion model for 3D point cloud generation on Hilbert-curve ordering and multifractal (topology-aware) conditioning signals to encode global structure; details beyond the abstract are unknown.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add HMC conditioning (VAE variant) on top of the DiT-3D bottleneck backbone | TopoDiT-3D · ShapeNet Car, 1-NNA-EMD | 1-NNA-EMD (distance from 50): 59.09 -> 49.81 | Improved |
| add HMC conditioning (VAE variant) | TopoDiT-3D · ShapeNet Chair, COV-CD | COV-CD: 56.67 -> 54.32 | Hurt |
| add HMC conditioning (bank variant) | TopoDiT-3D · ShapeNet Car, COV-EMD | COV-EMD: 51.98 -> 57.86 | Improved |
| add HMC conditioning (bank variant) | TopoDiT-3D · ShapeNet Chair, 1-NNA-CD | 1-NNA-CD: 48.87 -> 49.01 | No clear effect |
| add HMC conditioning (both variants) | TopoDiT-3D · ShapeNet Airplane | 1-NNA and COV: bank 1-NNA-CD 57.15 -> 54.79 | Improved |
| add HMC conditioning (both variants) | TopoDiT-3D · ShapeNet Lamp and Vessel | 1-NNA and COV: Vessel 1-NNA-EMD 78.96 -> 49.05 (bank) | Improved |
| remove HMC tokens (ablation) | HilbertGen-3D · ShapeNet Airplane ablation | 1-NNA-CD: 57.33 -> 55.57 when HMC tokens added | Improved |
| add global HMC embedding (ablation) | HilbertGen-3D · ShapeNet Airplane ablation | COV-CD: COV-CD 60.24 -> 60.98; 1-NNA-EMD 53.58 -> 53.66 | Mixed |
| add bottleneck fusion (ablation) | HilbertGen-3D · ShapeNet Airplane ablation | 1-NNA-CD: 60.00 -> 55.57 | Improved |
| add bottleneck fusion (ablation) | HilbertGen-3D · ShapeNet Airplane ablation | COV-CD: 55.92 -> 60.98 | Improved |
| use VAE-generated versus bank-sampled HMC conditions (ablation) | HilbertGen-3D · ShapeNet Chair, Airplane, Car | 1-NNA and COV | Mixed |
Closest papers by meaning
- SILSA: Sliding-Window Slice Latents for Topology-Preserving High-Resolution 3D Generation (NeurIPS 2026 · similarity 0.87)
- Ouroboros3D: Image-to-3D Generation via 3D-aware Recursive Diffusion (CVPR 2025 · similarity 0.87)
- GPLD3D: Latent Diffusion of 3D Shape Generative Models by Enforcing Geometric and Physical Priors (CVPR 2024 · similarity 0.87)
- Direct3D-S2: Gigascale 3D Generation Made Easy with Spatial Sparse Attention (NeurIPS 2025 · similarity 0.87)
- ShapeShifter: 3D Variations Using Multiscale and Sparse Point-Voxel Diffusion (CVPR 2025 · similarity 0.86)
- Scaling Diffusion Mamba with Bidirectional SSMs for Efficient 3D Shape Generation (AAAI 2025 · similarity 0.86)
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.
