Research map / Point clouds / Shape and surface completion
Implicit Filtering for Learning Neural Signed Distance Functions from 3D Point Clouds
Shengtao Li, Ge Gao, Yudong Liu, Ming Gu, Yu-Shen Liu
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 |
|---|---|---|---|
| add the zero level set filter loss L_zero to L_CD (ablation) | Implicit Filtering · FAMOUS | CD_L2 x1000: 0.031 -> 0.018 | Improved |
| add non-zero level set filtering L_field (ablation) | Implicit Filtering · FAMOUS | CD_L2 x1000: 0.018 -> 0.011 | Improved |
| add L_dist term (ablation) | Implicit Filtering · FAMOUS | CD_L2 x1000: 0.011 -> 0.008 | Improved |
| use bidirectional projection distance instead of single-direction (ablation) | Implicit Filtering · FAMOUS | F-score: 0.726 -> 0.778 | Improved |
| use Chamfer distance loss instead of pull loss (ablation) | Implicit Filtering · FAMOUS | CD_L2 x1000: 0.083 -> 0.031 | Improved |
| add Eikonal term alongside CD loss (ablation) | Implicit Filtering · FAMOUS | CD_L2 x1000: 0.008 -> 0.009 (CD_L2 without vs with Eikonal) | No clear effect |
| use Eikonal term instead of CD loss for gradient constraint (ablation) | Implicit Filtering · FAMOUS | CD_L2 x1000: 0.008 -> 0.021 | Hurt |
| increase CD loss weight alpha3 from 10 to 100 (ablation) | Implicit Filtering · FAMOUS | F-score: 0.778 -> 0.774 | Hurt |
| set alpha3 to 0 (drop CD constraint) (ablation) | Implicit Filtering · FAMOUS | F-score: 0.778 -> 0.758 | Hurt |
| widen sigma_n from 15 to 30/45/60 degrees (ablation) | Implicit Filtering · FAMOUS | F-score: 0.778 -> 0.771 / 0.764 / 0.767 | Hurt |
| set sigma_p to bounding box diagonal rather than max neighbor distance (ablation) | Implicit Filtering · FAMOUS | F-score: 0.778 -> 0.763 | Hurt |
| apply implicit filtering for scenes | Implicit Filtering vs PCP · 3D Scene | normal consistency: Lounge 0.934 (PCP) vs 0.926 (Ours) | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- DIGS — DiGS: Divergence Guided Shape Implicit Neural Representation for Unoriented Point Clouds (CVPR 2022) · better on 4 of 4 shared results
- GridPull — GridPull: Towards Scalability in Learning Implicit Representations from 3D Point Clouds (ICCV 2023) · better on 2 of 3 shared results
- NeuralPull — Neural-Pull: Learning Signed Distance Function from Point clouds by Learning to Pull Space onto Surface (ICML 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Few-Shot Unsupervised Implicit Neural Shape Representation Learning with Spatial Adversaries (ICML 2024 · similarity 0.91)
- Small Steps and Level Sets: Fitting Neural Surface Models with Point Guidance (CVPR 2024 · similarity 0.90)
- Inferring Neural Signed Distance Functions by Overfitting on Single Noisy Point Clouds through Finetuning Data-Driven based Priors (NeurIPS 2024 · similarity 0.89)
- MultiPull: Detailing Signed Distance Functions by Pulling Multi-Level Queries at Multi-Step (NeurIPS 2024 · similarity 0.89)
- Sharpening Neural Implicit Functions with Frequency Consolidation Priors (AAAI 2025 · similarity 0.89)
- LISR: Learning Linear 3D Implicit Surface Representation Using Compactly Supported Radial Basis Functions (AAAI 2024 · similarity 0.89)
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.
