Research map / Probabilistic and scientific modeling / Variational and generative sampling
MetaSDF: Meta-Learning Signed Distance Functions
Vincent Sitzmann, Eric Chan, Richard Tucker, Noah Snavely, Gordon Wetzstein
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 |
|---|---|---|---|
| use gradient-based meta-learning instead of auto-decoder latent optimization | shape-space learning · ShapeNet planes dense SDF | mean Chamfer distance: 0.073 -> 0.053 | Improved |
| use gradient-based meta-learning instead of auto-decoder latent optimization | shape-space learning · ShapeNet tables dense SDF | mean Chamfer distance: 0.553 -> 0.134 | Improved |
| use gradient-based meta-learning instead of auto-decoder latent optimization | shape-space learning · ShapeNet 3D reconstruction | inference time: 8 seconds to 0.4 seconds | Improved |
| use gradient-based meta-learning instead of PointNet set encoder | conditioning on zero-level set points · ShapeNet planes | mean Chamfer distance: 0.558 -> 0.139 | Improved |
| use gradient-based meta-learning instead of PointNet set encoder | conditioning on zero-level set points · ShapeNet tables | mean Chamfer distance: 0.693 -> 0.327 | Improved |
| use gradient-based meta-learning instead of latent-code optimization | 2D SDF inference · MNIST SDFs | inference time: 4 seconds -> 50 ms | Improved |
| replace clamped l1 loss with composite SDF + sign loss (ablation) | DeepSDF · ShapeNet planes | Chamfer distance | No clear effect |
| use composite SDF + sign loss instead of clamped l1 in inner loop (ablation) | MetaSDF · ShapeNet | training stability | Improved |
Closest papers by meaning
- Few ‘Zero Level Set’-Shot Learning of Shape Signed Distance Functions in Feature Space (ECCV 2022 · similarity 0.92)
- Deep Meta Functionals for Shape Representation (ICCV 2019 · similarity 0.89)
- Implicit Surface Representations As Layers in Neural Networks (ICCV 2019 · similarity 0.87)
- Implicit Geometric Regularization for Learning Shapes (ICML 2020 · similarity 0.87)
- PatchNets: Patch-Based Generalizable Deep Implicit 3D Shape Representations (ECCV 2020 · similarity 0.87)
- Few-Shot Unsupervised Implicit Neural Shape Representation Learning with Spatial Adversaries (ICML 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.
