Research map / 3D vision / Depth and stereo
Learning Transformation Synchronization
Xiangru Huang, Zhenxiao Liang, Xiaowei Zhou, Yao Xie, Leonidas J. Guibas, Qixing Huang
Check 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 |
|---|---|---|---|
| Use learned recurrent weighting and synchronization instead of FastGR pairwise alignments alone | FastGR alignments · Redwood test | mean rotation error: 34.1^{\circ} -> 6.9^{\circ} | Improved |
| Use learned recurrent weighting and synchronization instead of FastGR pairwise alignments alone | FastGR alignments · Redwood test | mean translation error: 0.58m -> 0.26m | Improved |
| Use learned recurrent weighting and synchronization instead of FastGR pairwise alignments alone | FastGR alignments · ScanNet test | mean rotation error: 68.8^{\circ} -> 42.9^{\circ} | Improved |
| Use learned recurrent weighting and synchronization instead of FastGR pairwise alignments alone | FastGR alignments · ScanNet test | mean translation error: 1.43m -> 1.16m | Improved |
| Replace RotAvg with learned recurrent weighting and synchronization | FastGR alignments · Redwood test | mean rotation error: 22.4^{\circ} -> 6.9^{\circ} | Improved |
| Replace RotAvg with learned recurrent weighting and synchronization | FastGR alignments · Redwood test | mean translation error: 0.418m; reduction: 39.0\% | Improved |
| Replace RotAvg with learned recurrent weighting and synchronization | FastGR alignments · ScanNet test | mean rotation error: reduction: 33.3\% | Improved |
| Replace RotAvg with learned recurrent weighting and synchronization | FastGR alignments · ScanNet test | mean translation error: reduction: 7.4\% | Improved |
| Jointly learn the weighting module rather than fixing its initial map weights (ablation) | Learning Transformation Synchronization · Redwood test | mean rotation error: 13.7^{\circ} -> 6.9^{\circ} | Improved |
| Jointly learn the weighting module rather than fixing its initial map weights (ablation) | Learning Transformation Synchronization · Redwood test | mean translation error: 0.255m -> 0.26m | Improved |
| Use recurrent weighting and synchronization rather than a single synchronization layer (ablation) | Learning Transformation Synchronization · Redwood and ScanNet | mean pose error | Improved |
Closest papers by meaning
- K-Best Transformation Synchronization (ICCV 2019 · similarity 0.88)
- Pose Synchronization Under Multiple Pair-Wise Relative Poses (CVPR 2023 · similarity 0.87)
- Dense Correspondences between Human Bodies via Learning Transformation Synchronization on Graphs (NeurIPS 2020 · similarity 0.86)
- Local-consistent Transformation Learning for Rotation-invariant Point Cloud Analysis (CVPR 2024 · similarity 0.85)
- Neural Isometries: Taming Transformations for Equivariant ML (NeurIPS 2024 · similarity 0.85)
- SyncNoise: Geometrically Consistent Noise Prediction for Instruction-based 3D Editing (AAAI 2025 · similarity 0.84)
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.
