Research map / Point clouds / Shape and surface completion
Optimization over Disentangled Encoding: Unsupervised Cross-Domain Point Cloud Completion via Occlusion Factor Manipulation
Jingyu Gong, Fengqi Liu, Jiachen Xu, Min Wang, Xin Tan, Zhizhong Zhang, Ran Yi, Haichuan Song, Yuan Xie, Lizhuang Ma
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 disentangled representation learning (domain, shape, occlusion factors) (ablation) | OptDE baseline · ModelNet (avg) | Chamfer distance: 30.19 -> 17.84 | Improved |
| add factor permutation consistency (ablation) | DE · ModelNet (avg) | Chamfer distance: 17.84 -> 16.96 | Improved |
| add optimization stage over disentangled encoding (ablation) | DE · ModelNet (avg) | Chamfer distance: 16.96 -> 15.94 | Improved |
| add optimization stage (ablation) | DE · 3D-FUTURE (avg) | Chamfer distance: 29.42 -> 28.99 | Improved |
| add optimization stage (ablation) | DE · ScanNet Chair (real) | UCD (x1e4): 2.8 -> 2.6 | Improved |
| add optimization stage (ablation) | DE · ScanNet Chair (real) | UHD (x1e2): 5.4 -> 5.5 | Hurt |
| add optimization stage (ablation) | DE · ModelNet Car | Chamfer distance: 9.80 -> 9.80 | No clear effect |
| use OptDE instead of ShapeInversion | ShapeInversion · ModelNet (avg) | Chamfer distance: 41.61 -> 15.94 | Improved |
| use OptDE instead of Cycle4Completion | Cycle4Completion · 3D-FUTURE (avg) | Chamfer distance: 45.39 -> 28.99 | Improved |
| good initialization from disentangled encoding instead of ShapeInversion optimization | OptDE · ModelNet optimization speed | time: 23.56s for 800 iterations -> 0.12s for 4 iterations | Improved |
| train with target-domain paired data (oracle) | OptDE · ModelNet (avg) | Chamfer distance: 15.94 -> 13.07 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Cycle4Completion — Cycle4Completion: Unpaired Point Cloud Completion Using Cycle Transformation With Missing Region Coding (CVPR 2021) · better on 8 of 8 shared results
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
- Learning a Structured Latent Space for Unsupervised Point Cloud Completion (CVPR 2022 · similarity 0.90)
- Cross-modal Learning for Image-Guided Point Cloud Shape Completion (NeurIPS 2022 · similarity 0.89)
- Complete Structure Guided Point Cloud Completion via Cluster- and Instance-Level Contrastive Learning (NeurIPS 2025 · similarity 0.89)
- Point Cloud Completion by Skip-Attention Network With Hierarchical Folding (CVPR 2020 · similarity 0.88)
- CDPNet: Cross-Modal Dual Phases Network for Point Cloud Completion (AAAI 2024 · similarity 0.88)
- Orthogonal Dictionary Guided Shape Completion Network for Point Cloud (AAAI 2024 · 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.
