Research map / Point clouds / Shape and surface completion
ShapeFormer: Transformer-Based Shape Completion via Sparse Representation
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 ShapeFormer with VQDIF rather than the strongest listed baseline | ShapeNet high scan ambiguity · ShapeNet high scan ambiguity | FPD: 2.55 -> 1.45 | Helped |
| Use ShapeFormer rather than IF-Net | ShapeNet low scan ambiguity · ShapeNet low scan ambiguity | CD: 0.79 -> 0.74 | Helped |
| Use ShapeFormer rather than IF-Net | ShapeNet low scan ambiguity · ShapeNet low scan ambiguity | F1: 73.8 -> 70.3 | Hurt |
| Use ShapeFormer rather than cGAN | PartNet completion · PartNet (Table, Chair, Lamp average) | TMD: 3.05 -> 3.96 | Helped |
| Use ShapeFormer rather than cGAN | PartNet completion · PartNet (Table, Chair, Lamp average) | MMD: 1.98 -> 1.32 | Helped |
| Use ShapeFormer rather than SInv. | PartNet completion · PartNet (Table, Chair, Lamp average) | UHD: 2.32 -> 0.98 | Helped |
| Use ShapeFormer rather than cGAN | PartNet completion · PartNet (Table, Chair, Lamp average) | FPD: 2.95 -> 1.22 | Helped |
| Use top-p=0 sampling instead of top-p=0.4 (ablation) | ShapeFormer · ShapeNet low scan ambiguity | F1: 70.3 -> 71.4 | Helped |
| Use top-p=0 sampling instead of top-p=0.4 (ablation) | ShapeFormer · ShapeNet high scan ambiguity | FPD: 1.45 -> 1.83 | Hurt |
| Increase VQDIF feature resolution from 8 to 16 (ablation) | VQDIF · ShapeNet auto-encoding | CD: 1.90 -> 0.98 | Helped |
| Increase VQDIF feature resolution from 16 to 32 (ablation) | VQDIF · ShapeNet auto-encoding | F1: 88.1 -> 96.4 | Helped |
| Increase VQDIF feature resolution from 16 to 32 (ablation) | VQDIF · ShapeNet auto-encoding | sequence length: 217 -> 889 | Hurt |
Closest papers by meaning
- Shape Completion Using 3D-Encoder-Predictor CNNs and Shape Synthesis (CVPR 2017 · similarity 0.90)
- Deformable Shape Completion With Graph Convolutional Autoencoders (CVPR 2018 · similarity 0.89)
- FSC: Few-point Shape Completion (CVPR 2024 · similarity 0.88)
- RaySt3R: Predicting Novel Depth Maps for Zero-Shot Object Completion (NeurIPS 2025 · similarity 0.87)
- PCDreamer: Point Cloud Completion Through Multi-view Diffusion Priors (CVPR 2025 · similarity 0.87)
- Sparc3D: Sparse Representation and Construction for High-Resolution 3D Shapes Modeling (NeurIPS 2025 · similarity 0.87)
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.