Research map / Point clouds / Shape and surface completion
Grad-PU: Arbitrary-Scale Point Cloud Upsampling via Gradient Descent With Learned Distance Functions
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 |
|---|---|---|---|
| use point-to-point distance regression instead of coordinate-offset prediction (ablation) | Grad-PU · PU1K, 4× | CD (10^-3): 1.170 -> 0.404; 0.663 -> 0.404 | Helped |
| use midpoint-interpolated points for both the network input and initialization (ablation) | Grad-PU · PU1K, 4× | CD (10^-3): 0.713 -> 0.404; 0.426 -> 0.404; 0.654 -> 0.404 | Helped |
| use P3DConv rather than EdgeConv for feature extraction (ablation) | Grad-PU · PU1K, 4× | CD (10^-3): 0.878 -> 0.404 | Helped |
| include both local and global features in the distance regressor (ablation) | Grad-PU · PU1K, 4× | CD (10^-3): 0.729 -> 0.404; 0.423 -> 0.404 | Helped |
| add Gaussian noise jitter to training points (ablation) | Grad-PU · PU1K, 4× | CD (10^-3): 0.520 -> 0.404 | Helped |
| use gradient descent on the predicted distance rather than distance-scaled normalized-gradient projection (ablation) | Grad-PU · PU1K, 4× | CD (10^-3): 0.872 -> 0.404 | Helped |
| use Grad-PU instead of the strongest listed baseline | PU-GAN, 4× · PU-GAN, 4× | CD (10^-3): 0.259 -> 0.245 | Helped |
| use Grad-PU instead of the strongest listed baseline | PU-GAN, 16× · PU-GAN, 16× | CD (10^-3): 0.152 -> 0.108 | Helped |
| use Grad-PU instead of PU-Transformer | PU1K · PU1K, 4× | P2F (10^-3): 1.277 -> 1.474 | Hurt |
| use Grad-PU for arbitrary-scale upsampling rather than NePs | PU-GAN · PU-GAN, 6× and 7× | P2F (10^-3): 1.809 -> 1.981; 1.795 -> 1.940 | Mixed |
Closest papers by meaning
- SPU-IMR: Self-supervised Arbitrary-scale Point Cloud Upsampling via Iterative Mask-recovery Network (AAAI 2025 · similarity 0.92)
- Learning Continuous Implicit Field with Local Distance Indicator for Arbitrary-Scale Point Cloud Upsampling (AAAI 2024 · similarity 0.91)
- Arbitrary-Scale Point Cloud Upsampling by Voxel-Based Network with Latent Geometric-Consistent Learning (AAAI 2024 · similarity 0.90)
- A Conditional Denoising Diffusion Probabilistic Model for Point Cloud Upsampling (CVPR 2024 · similarity 0.90)
- RepKPU: Point Cloud Upsampling with Kernel Point Representation and Deformation (CVPR 2024 · similarity 0.88)
- PLATYPUS: Progressive Local Surface Estimator for Arbitrary-Scale Point Cloud Upsampling (AAAI 2025 · 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.