Research map / Point clouds / Shape and surface completion
Learning Canonical Embeddings for Unsupervised Shape Correspondence With Locally Linear Transformations
Pan Li He, Patrick Emami, Sanjay Ranka, Anand Rangarajan
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 the mapping smoothness loss E_r to DPC (ablation) | DPC · SURREAL/SHREC | accuracy: 11.4% -> 17.7% | Improved |
| add the mapping smoothness loss E_r to LTENet (ablation) | LTENet · SURREAL/SHREC | accuracy: 20.6% -> 20.7% | No clear effect |
| add cross-reconstruction to self-reconstruction (ablation) | LTENet · SURREAL/SHREC | accuracy: 2.3% -> 20.6% | Improved |
| add self-reconstruction to cross-reconstruction (ablation) | LTENet · SURREAL/SHREC | error: 6.5 -> 5.9 | Improved |
| add E_r with lambda_reg=10 to self and cross losses (ablation) | LTENet · SURREAL/SHREC | error: 5.9 -> 5.8 | Improved |
| use CS divergence instead of Chamfer distance (ablation) | LTENet · SURREAL/SHREC | accuracy: 13.9% -> 20.7% | Improved |
| use CS divergence instead of Earth mover's distance (ablation) | LTENet · SURREAL/SHREC | accuracy: 19.6% -> 20.7% | Improved |
| use LTENet instead of DPC | correspondence learning · SURREAL/SHREC | accuracy: 17.7% -> 20.7% | Improved |
| use LTENet instead of DPC | correspondence learning · SMAL/TOSCA | accuracy: 33.2% -> 38.1% | Improved |
| use LTENet instead of DPC | correspondence learning · SHREC/SHREC | accuracy: 15.3% -> 14.3% | No clear effect |
| apply optimal global linear transformation to learned embeddings (SMAL/TOSCA) (ablation) | DPC embeddings · SMAL/TOSCA | accuracy | Hurt |
| apply optimal global linear transformation to learned embeddings (ablation) | LTENet · SURREAL/SHREC and SMAL/TOSCA | accuracy | 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.
- CorrNet3D — CorrNet3D: Unsupervised End-to-End Learning of Dense Correspondence for 3D Point Clouds (CVPR 2021) · better on 5 of 5 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
- Correspondence learning via linearly-invariant embedding (NeurIPS 2020 · similarity 0.91)
- CorrNet3D: Unsupervised End-to-End Learning of Dense Correspondence for 3D Point Clouds (CVPR 2021 · similarity 0.89)
- Learning Implicit Functions for Dense 3D Shape Correspondence of Generic Objects (TPAMI 2023 · similarity 0.89)
- Learning SO(3)-Invariant Semantic Correspondence via Local Shape Transform (CVPR 2024 · similarity 0.89)
- Deep Shells: Unsupervised Shape Correspondence with Optimal Transport (NeurIPS 2020 · similarity 0.88)
- Adaptive Siamese Masked Autoencoder with Global Optimization for Unsupervised Point Cloud Shape Correspondence (AAAI 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.
