Research map / Point clouds / Shape and surface completion
Point Cloud Completion by Skip-Attention Network With Hierarchical Folding
Check 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 SA-Net instead of TopNet | point cloud completion · ShapeNet completion, average | per-point CD (×10^4): 9.72 -> 7.74 | Helped |
| remove skip-attention (ablation) | SA-Net · ShapeNet completion, Plane | CD (×10^4): 2.18 -> 2.31 | Hurt |
| replace cosine skip-attention with learnable attention (ablation) | SA-Net · ShapeNet completion, Plane | CD (×10^4): 2.18 -> 2.25 | Hurt |
| replace learnable folding-block self-attention with cosine similarity (ablation) | SA-Net · ShapeNet completion, Plane | CD (×10^4): 2.18 -> 2.34 | Hurt |
| use cosine skip-attention rather than learnable skip-attention (ablation) | SA-Net · ShapeNet completion, Plane | CD (×10^4): 2.25 -> 2.18 | Helped |
| use learnable self-attention rather than cosine similarity (ablation) | SA-Net folding block · ShapeNet completion, Plane | CD (×10^4): 2.34 -> 2.18 | Helped |
| train with EMD loss only rather than the combined losses (ablation) | SA-Net · ShapeNet completion, Plane | CD (×10^4): 2.18 -> 2.39 | Hurt |
| train with CD loss only rather than the combined losses (ablation) | SA-Net · ShapeNet completion, Plane | EMD (×10^2): 3.02 -> 4.58 | Hurt |
| reduce input from 2048 to 1024 points (ablation) | SA-Net · ShapeNet completion, Plane | CD (×10^4): 2.18 -> 2.28 | Hurt |
| reduce input from 2048 to 512 points (ablation) | SA-Net · ShapeNet completion, Plane | CD (×10^4): 2.18 -> 2.45 | Hurt |
| reduce input from 2048 to 256 points (ablation) | SA-Net · ShapeNet completion, Plane | CD (×10^4): 2.18 -> 3.31 | Hurt |
| add skip-attention to PointNet++ interpolation layers | PointNet++ semantic segmentation · ShapeNet semantic segmentation | mpIoU (%): 81.9 -> 83.0 | Helped |
Closest papers by meaning
- DC-PCN: Point Cloud Completion Network with Dual-Codebook Guided Quantization (AAAI 2025 · similarity 0.90)
- PointAttN: You Only Need Attention for Point Cloud Completion (AAAI 2024 · similarity 0.90)
- DeCoTR: Enhancing Depth Completion with 2D and 3D Attentions (CVPR 2024 · similarity 0.90)
- Orthogonal Dictionary Guided Shape Completion Network for Point Cloud (AAAI 2024 · similarity 0.89)
- PointCFormer: A Relation-Based Progressive Feature Extraction Network for Point Cloud Completion (AAAI 2025 · similarity 0.89)
- SymmCompletion: High-Fidelity and High-Consistency Point Cloud Completion with Symmetry Guidance (AAAI 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.