Research map / 3D vision / Face and shape
Weakly-supervised 3D Shape Completion in the Wild
Jiayuan Gu, Wei-Chiu Ma, Sivabalan Manivasagam, Wenyuan Zeng, Zihao Wang, Yuwen Xiong, Hao Su, Raquel Urtasun
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 shape-projection-matching-observation term (ablation) | Ours · ShapeNet Airplane | chamfer distance x100: 2.80 -> 1.65 | Improved |
| add shape-projection-matching-observation term (ablation) | Ours · ShapeNet Car | chamfer distance x100: 2.96 -> 2.68 | Improved |
| add shape-projection-matching-observation term (ablation) | Ours · ShapeNet Chair | chamfer distance x100: 3.94 -> 3.33 | Improved |
| add hindsight loss (ablation) | Ours · ShapeNet Airplane | chamfer distance x100: 2.32 -> 1.65 | Improved |
| add hindsight loss (ablation) | Ours · ShapeNet Car | chamfer distance x100: 2.82 -> 2.68 | Improved |
| add hindsight loss (ablation) | Ours · ShapeNet Chair | chamfer distance x100: 3.80 -> 3.33 | Improved |
| add projection term (3D vehicle dataset) (ablation) | Ours · 3D vehicle dataset | precision/coverage/chamfer distance | Mixed |
| increase training views per instance from 2 to 8 (ablation) | Ours · 3D vehicle dataset | chamfer distance: 0.307 -> 0.242 | Improved |
| increase training views per instance from 2 to 4 (ablation) | Ours · 3D vehicle dataset | rotation error: 6.185 -> 4.208 | Improved |
| use weakly supervised canonical-pose method instead of DPC | DPC · ShapeNet Airplane registration | pose accuracy: 74.17 -> 92.87 | Improved |
| use 2D-projected input/output loss variant for cars | Ours · ShapeNet Car registration | pose accuracy: 84.75 -> 91.03 | Improved |
| evaluate against partial GT rather than symmetric complete GT | Ours · 3D vehicle dataset | chamfer distance: 0.255 -> 0.195 | No clear effect |
Closest papers by meaning
- Learning 3D Shape Completion Under Weak Supervision (IJCV 2018 · similarity 0.89)
- Self-supervised Shape Completion via Involution and Implicit Correspondences (ECCV 2024 · similarity 0.88)
- Learning 3D Shape Completion From Laser Scan Data With Weak Supervision (CVPR 2018 · similarity 0.88)
- ShapeFormer: Transformer-Based Shape Completion via Sparse Representation (CVPR 2022 · similarity 0.85)
- FSC: Few-point Shape Completion (CVPR 2024 · similarity 0.83)
- Learning Generalizable Shape Completion with SIM(3) Equivariance (NeurIPS 2025 · similarity 0.83)
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.
