Research map / 3D vision / Face and shape
Single Depth-image 3D Reflection Symmetry and Shape Prediction
Zhaoxuan Zhang, Bo Dong, Tong Li, Felix Heide, Pieter Peers, Baocai Yin, Xin Yang
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 symmetry plane alignment (vs random rotation and perturbed center) (ablation) | ISCNet · ShapeNet (8 categories) | Chamfer Distance x10^2: 2.04 -> 1.48 | Improved |
| process symmetric viewpoints as pairs in inpainting (ablation) | ISCNet · ShapeNet (8 categories) | Chamfer Distance x10^2: 1.82 -> 1.48 | Improved |
| add inpainted normal maps as input to depth inpainting (ablation) | ISCNet · ShapeNet (8 categories) | Chamfer Distance x10^2: 1.94 -> 1.48 | Improved |
| increase iterations from 4 to 20 (ablation) | ISCNet · ShapeNet (8 categories) | Chamfer Distance x10^2: 1.61 -> 1.48 | Improved |
| iteratively refine symmetry plane (ablation) | ISCNet · symmetry plane estimation (error < 5 deg) | fraction of estimates: 32.86% -> 56.96% | Improved |
| use ISCNet instead of Front2Back with ground-truth symmetry | ISCNet · ShapeNet (8 categories) | Chamfer Distance x10^2: 1.91 -> 1.48 | Improved |
| use ISCNet on unseen classes without retraining | ISCNet · ShapeNet unseen-5 classes | Chamfer Distance x10^2: 2.76 -> 2.11 (vs Front2Back) | Mixed |
| use ISCNet instead of SnowFlakeNet on plane shapes | ISCNet · ShapeNet plane | Chamfer Distance x10^2: 1.08 -> 0.82 | Hurt |
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.
- Front2Back — Front2Back: Single View 3D Shape Reconstruction via Front to Back Prediction (CVPR 2020) · 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
- Learning to Detect 3D Symmetry From Single-View RGB-D Images With Weak Supervision (TPAMI 2022 · similarity 0.88)
- Symmetry Strikes Back: From Single-Image Symmetry Detection to 3D Generation (CVPR 2025 · similarity 0.88)
- NeRD: Neural 3D Reflection Symmetry Detector (CVPR 2021 · similarity 0.88)
- Front2Back: Single View 3D Shape Reconstruction via Front to Back Prediction (CVPR 2020 · similarity 0.87)
- Learning Generalizable Shape Completion with SIM(3) Equivariance (NeurIPS 2025 · similarity 0.86)
- CDPNet: Cross-Modal Dual Phases Network for Point Cloud Completion (AAAI 2024 · similarity 0.85)
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.
