Research map / Point clouds / Shape and surface completion
Robustifying Generalizable Implicit Shape Networks with a Tunable Non-Parametric Model
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 |
|---|---|---|---|
| replace Euclidean-space regression (Fourier encoding) with regression in network feature space (ablation) | Ours (m=500) · ShapeNet Lamp | IoU: 0.45 -> 0.87 | Helped |
| learn kernel hyperparameters theta by gradient descent (ablation) | Ours · ShapeNet Lamp | IoU: 0.86 -> 0.87 | Helped |
| learn Nystrom samples Z by gradient descent (ablation) | Ours · ShapeNet Lamp | CD1: 0.53 -> 0.42 | Helped |
| add hyperparameter tuning (vs no tuning, m=1k) (ablation) | Nystrom KRR · ShapeNet Lamp | CD1: 0.61 -> 0.42 | Helped |
| increase Nystrom samples without tuning from 1k to 10k (ablation) | Nystrom KRR w/o tune · ShapeNet Lamp | CD1: 0.61 -> 0.53 | Helped |
| increase Nystrom samples with tuning from 500 to 1000 (ablation) | Ours · ShapeNet Lamp | IoU: 0.87 -> 0.88 | Helped |
| increase Nystrom samples with tuning from 1000 to 5000 (ablation) | Ours · ShapeNet Lamp | IoU: 0.88 -> 0.86 | Hurt |
| apply KRR adaptation to Poco at 10k input points (ablation) | Poco · ShapeNet Lamp | IoU: 0,80 -> 0,87 | Helped |
| apply KRR adaptation to Poco at 500 input points (ablation) | Poco · ShapeNet Lamp | IoU: 0,71 -> 0,73 | Helped |
| increase input size from 3k to 10k (ablation) | Poco · ShapeNet Lamp | IoU: 0,85 -> 0,80 | Hurt |
| test-time finetuning of Conv (SAC) | Conv · ShapeNet | IoU: 0,84 -> 0,81 | Mixed |
| test-time finetuning of Conv (SAC) on Faust | Conv · Faust | IoU: 0,85 -> 0,77 | Mixed |
Closest papers by meaning
- Implicit Geometric Regularization for Learning Shapes (ICML 2020 · similarity 0.89)
- Neural Fields As Learnable Kernels for 3D Reconstruction (CVPR 2022 · similarity 0.88)
- Few-Shot Unsupervised Implicit Neural Shape Representation Learning with Spatial Adversaries (ICML 2024 · similarity 0.86)
- Enhancing Implicit Shape Generators Using Topological Regularizations (ICML 2024 · similarity 0.86)
- NeuralIndicator: Implicit Surface Reconstruction from Neural Indicator Priors (ICML 2024 · similarity 0.85)
- IPoD: Implicit Field Learning with Point Diffusion for Generalizable 3D Object Reconstruction from Single RGB-D Images (CVPR 2024 · similarity 0.85)
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.