Research map / Point clouds / Shape and surface completion
SeedFormer: Patch Seeds Based Point Cloud Completion with Upsample Transformer
Haoran Zhou, Yun Cao, Wenqing Chu, Junwei Zhu, Tong Lu, Ying Tai, Chengjie Wang
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 Patch Seeds representation instead of global feature (ablation) | SeedFormer · PCN dataset | CD-Avg: 6.97 -> 6.74 | Improved |
| use 256 seeds instead of 128 (ablation) | SeedFormer · PCN dataset | CD-Avg: 7.00 -> 6.74 | Improved |
| use 512 seeds instead of 256 (ablation) | SeedFormer · PCN dataset | CD-Avg: 6.74 -> 6.75 | No clear effect |
| use Upsample Transformer instead of folding generator (ablation) | SeedFormer · PCN dataset | CD-Avg: 6.93 -> 6.74 | Improved |
| use Upsample Transformer instead of deconvolution generator (ablation) | SeedFormer · PCN dataset | CD-Avg: 6.90 -> 6.74 | Improved |
| use Upsample Transformer instead of graph convolution (ablation) | SeedFormer · PCN dataset | CD-Avg: 6.88 -> 6.74 | Improved |
| use Upsample Transformer instead of point-wise attention (ablation) | SeedFormer · PCN dataset | CD-Avg: 6.85 -> 6.74 | Improved |
| remove softmax in seed generator self-attention (ablation) | SeedFormer · PCN dataset | CD-Avg: 6.83 -> 6.74 | Improved |
| use scaled-softmax instead of no softmax (ablation) | SeedFormer · PCN dataset | CD-Avg: 6.74 -> 6.80 | Hurt |
| add positional encoding (ablation) | Upsample Transformer · PCN dataset | CD-Avg: 6.88 -> 6.80 | Improved |
| add regional encoding from seed features to positional encoding (ablation) | Upsample Transformer · PCN dataset | CD-Avg: 6.80 -> 6.74 | Improved |
| combine positional and regional into one joint encoding (ablation) | Upsample Transformer · PCN dataset | CD-Avg: 6.74 -> 6.78 | Hurt |
Later papers that compared with this method
18 later accepted papers report SeedFormer as a baseline in their results tables; 18 of them report a better number on most of the results they share with it.
- Hyper-PCN: Hypergraph-Based Point Cloud Completion via High-Order Correlation Modeling (CVPR 2026 · better on 4 of 4 shared results)
- SplAttN: Bridging 2D and 3D with Gaussian Soft Splatting and Attention for Point Cloud Completion (ICML 2026 · better on 3 of 3 shared results)
- Semantic Guided Part Relation-aware Network for Point Cloud Completion (AAAI 2026 · better on 2 of 2 shared results)
- PQDT: Pseudo-Query Dual Transformer for Robust Point Cloud Restoration (CVPR 2026 · better on 1 of 1 shared result)
- Simba: Towards High-Fidelity and Geometrically-Consistent Point Cloud Completion via Transformation Diffusion (AAAI 2026 · better on 1 of 1 shared result)
- ESCAPE: Equivariant Shape Completion via Anchor Point Encoding (CVPR 2025 · better on 2 of 2 shared results)
- PointSea: Point Cloud Completion via Self-structure Augmentation (IJCV 2025 · better on 2 of 2 shared results)
- PointAttN: You Only Need Attention for Point Cloud Completion (AAAI 2024 · better on 3 of 5 shared results)
10 more not listed.
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 4 of 4 on most shared results.
- PoinTr — PoinTr: Diverse Point Cloud Completion With Geometry-Aware Transformers (ICCV 2021) · better on 6 of 6 shared results
- GRNet — GRNet: Gridding Residual Network for Dense Point Cloud Completion (ECCV 2020) · better on 3 of 3 shared results
- PMP-Net — PMP-Net: Point Cloud Completion by Learning Multi-Step Point Moving Paths (CVPR 2021) · better on 1 of 1 shared result
- SnowflakeNet — SnowflakeNet: Point Cloud Completion by Snowflake Point Deconvolution With Skip-Transformer (ICCV 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
- Orthogonal Dictionary Guided Shape Completion Network for Point Cloud (AAAI 2024 · similarity 0.90)
- AnchorFormer: Point Cloud Completion From Discriminative Nodes (CVPR 2023 · similarity 0.90)
- PointCFormer: A Relation-Based Progressive Feature Extraction Network for Point Cloud Completion (AAAI 2025 · similarity 0.88)
- PCDreamer: Point Cloud Completion Through Multi-view Diffusion Priors (CVPR 2025 · similarity 0.88)
- GenPC: Zero-shot Point Cloud Completion via 3D Generative Priors (CVPR 2025 · similarity 0.88)
- SymmCompletion: High-Fidelity and High-Consistency Point Cloud Completion with Symmetry Guidance (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.
