Research map / 3D vision / Face and shape
Neural Geometric Level of Detail: Real-Time Rendering With Implicit 3D Shapes
Check 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 |
|---|---|---|---|
| Represent geometry with a sparse octree feature volume and small LOD-specific decoders | Neural geometric LOD · ShapeNet150 | Chamfer-L1: 0.077 -> 0.062 | Helped |
| Represent geometry with a sparse octree feature volume and small LOD-specific decoders | Neural geometric LOD · Thingi32 | Chamfer-L1: 0.0329 -> 0.0271 | Helped |
| Use the sparse CUDA renderer instead of the baseline sphere-tracing implementation | Neural geometric LOD · V Mech, 1920 × 1080 | frametime (ms): 10843 -> 91 | Helped |
| Increase the octree representation from LOD 4 to LOD 5 (ablation) | Neural geometric LOD · Thingi32 | Chamfer-L1: 0.0273 -> 0.0271 | Helped |
| Increase the octree representation from LOD 5 to LOD 6 (ablation) | Neural geometric LOD · TurboSquid16 | Chamfer-L1: 0.085 -> 0.076 | Helped |
| Train for 30 epochs rather than 100 epochs (ablation) | Neural geometric LOD · Thingi32 | Chamfer-L1: 0.0271 -> 0.0278 | Helped |
| Initialize from pretrained single-shape weights before 30-epoch training (ablation) | Neural geometric LOD · Thingi32 | Chamfer-L1: 0.0278 -> 0.0279 | No clear effect |
| Freeze the surface extractor pretrained on one shape and train only the feature volume (ablation) | Neural geometric LOD · Thingi32 | Chamfer-L1: 0.0271 -> 0.0272 | No clear effect |
| Reduce feature dimension to 8 (ablation) | Neural geometric LOD · SDF reconstruction | reconstruction quality | No clear effect |
| Use progressive training that adds lower levels during training (ablation) | Neural geometric LOD on TurboSquid16 · TurboSquid16 | reconstruction quality | Helped |
Closest papers by meaning
- Neural Signed Distance Function Inference through Splatting 3D Gaussians Pulled on Zero-Level Set (NeurIPS 2024 · similarity 0.90)
- LTM: Lightweight Textured Mesh Extraction and Refinement of Large Unbounded Scenes for Efficient Storage and Real-time Rendering (CVPR 2024 · similarity 0.89)
- DUDF: Differentiable Unsigned Distance Fields with Hyperbolic Scaling (CVPR 2024 · similarity 0.89)
- Few-Shot Unsupervised Implicit Neural Shape Representation Learning with Spatial Adversaries (ICML 2024 · similarity 0.89)
- Unsigned Orthogonal Distance Fields: An Accurate Neural Implicit Representation for Diverse 3D Shapes (CVPR 2024 · similarity 0.89)
- Small Steps and Level Sets: Fitting Neural Surface Models with Point Guidance (CVPR 2024 · similarity 0.88)
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.