Research map / 3D vision / Face and shape
3DILG: Irregular Latent Grids for 3D Generative Modeling
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 number of latents M from 64 to 512 (ablation) | 3DILG · ShapeNet-v2 reconstruction (all categories) | IoU: 0.923 -> 0.953 | Helped |
| increase number of latents M from 64 to 512 (ablation) | 3DILG · ShapeNet-v2 reconstruction (all categories) | Chamfer-L1: 0.048 -> 0.040 | Helped |
| add vector quantization (ablation) | 3DILG M=512 · ShapeNet-v2 reconstruction (all categories) | IoU: 0.953 -> 0.950 | Hurt |
| add vector quantization (ablation) | 3DILG M=64 · ShapeNet-v2 reconstruction (all categories) | IoU: 0.923 -> 0.907 | Hurt |
| replace regular latent grid with irregular latents | ConvONet · ShapeNet-v2 reconstruction (all categories) | IoU: 0.888 -> 0.923 at M=64 | Helped |
| replace multiscale grid IF-Net with irregular latents M=512 | IF-Net · ShapeNet-v2 reconstruction (all categories) | IoU: 0.934 -> 0.953 | Helped |
| replace IF-Net with 3DILG M=512 | IF-Net · ShapeNet-v2 reconstruction (all categories) | F-Score: 0.967 -> 0.966 | No clear effect |
| replace 8^3 regular latent grid with irregular latents for category-conditioned generation | autoregressive transformer · Category-conditioned generation, ShapeNet mean over 7 categories | FID: 67.641 -> 57.617 | Helped |
Closest papers by meaning
- Make-A-Shape: a Ten-Million-scale 3D Shape Model (ICML 2024 · similarity 0.88)
- Mosaic-SDF for 3D Generative Models (CVPR 2024 · similarity 0.88)
- GPLD3D: Latent Diffusion of 3D Shape Generative Models by Enforcing Geometric and Physical Priors (CVPR 2024 · similarity 0.88)
- ShapeShifter: 3D Variations Using Multiscale and Sparse Point-Voxel Diffusion (CVPR 2025 · similarity 0.88)
- Enhancing Implicit Shape Generators Using Topological Regularizations (ICML 2024 · similarity 0.87)
- GALA: Geometry-Aware Local Adaptive Grids for Detailed 3D Generation (ICLR 2025 · 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.