Research map / Point clouds / Shape and surface completion
Deep Implicit Surface Point Prediction Networks
Rahul Venkatesh, Tejan Karmali, Sarthak Sharma, Aurobrata Ghosh, R. Venkatesh Babu, László A. Jeni, Maneesh Singh
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 projection step to sphere tracing (ablation) | CSP sphere tracing · ShapeNet | Depth error (MAE): 0.016 -> 0.014 | Improved |
| use Jacobian normals instead of forward-mode normals at alpha=0 | CSP · ShapeNet | Normal map similarity: 0.620 -> 0.913 | Improved |
| step back along ray by alpha=0.005 for forward-mode normals (ablation) | CSP (fwd.) · ShapeNet | Normal map similarity: 0.620 -> 0.912 | Improved |
| increase step-back alpha from 0.005 to 0.05 for Jacobian normals (ablation) | CSP (jac.) · ShapeNet | Normal map similarity: 0.920 -> 0.871 | Hurt |
| use CSP forward normals instead of NDF gradient normals | CSP (fwd.) · Rendering 512x512 image | Rendering time: 0.063s -> 0.003s | Improved |
| use CSP instead of NDF on unprocessed shapes | CSP · ShapeNet unprocessed | Normal cosine similarity: 0.86 -> 0.92 | Improved |
| use CSP instead of NDF on unprocessed shapes | CSP · ShapeNet unprocessed | Chamfer-L2 (x10^4): 1.73 -> 1.28 | Improved |
| use CSP instead of SAL on unprocessed shapes | CSP · ShapeNet unprocessed | Depth error: 0.025 -> 0.014 | Improved |
| use CSP instead of NDF on closed shapes | CSP · ShapeNet closed shapes | Chamfer-L2 (x10^4): 0.05 -> 0.1 | Hurt |
| use per-class fine-tuned CSP instead of DeepSDF | CSP · ShapeNet Sofas | Chamfer-L2 (x10^4): 3.29 -> 0.32 | Improved |
| use coarse-to-fine meshing instead of NDF point cloud plus BPA | CSP meshing · representative ShapeNet example | Meshing time: 104s -> 2.50s | Improved |
| use coarse-to-fine meshing instead of dense evaluation | CSP meshing · representative ShapeNet example | Function evaluations: 12M -> 636k | Improved |
Closest papers by meaning
- Neural Unsigned Distance Fields for Implicit Function Learning (NeurIPS 2020 · similarity 0.91)
- Learning Consistency-Aware Unsigned Distance Functions Progressively from Raw Point Clouds (NeurIPS 2022 · similarity 0.91)
- DUDF: Differentiable Unsigned Distance Fields with Hyperbolic Scaling (CVPR 2024 · similarity 0.89)
- LISR: Learning Linear 3D Implicit Surface Representation Using Compactly Supported Radial Basis Functions (AAAI 2024 · similarity 0.89)
- Small Steps and Level Sets: Fitting Neural Surface Models with Point Guidance (CVPR 2024 · similarity 0.89)
- A Lightweight UDF Learning Framework for 3D Reconstruction Based on Local Shape Functions (CVPR 2025 · 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.
