Research map / Label-efficient and robust learning / Recognition and long-tail
Learning With Batch-Wise Optimal Transport Loss for 3D Shape Recognition
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 batch-wise optimal transport weighting instead of pair-wise metric learning | 3D shape retrieval · SHREC13 | mAP: 0.752 -> 0.754 | Helped |
| use batch-wise optimal transport weighting instead of pair-wise metric learning | 3D shape retrieval · SHREC14 | mAP: 0.401 -> 0.591 | Helped |
| use the proposed batch-wise method rather than the strongest listed baseline | 3D shape classification · ModelNet10 | accuracy: 92.3% -> 93.7 % | Helped |
| use the proposed batch-wise method rather than the strongest listed baseline | 3D shape retrieval · ModelNet40 | mAP: 83.5% -> 83.8 % | Helped |
| use the proposed batch-wise method rather than the strongest listed baseline | 3D shape classification · ModelNet40 | accuracy: 90.7 % -> 90.3 % | Hurt |
| add optimal transport-derived importance weights rather than uniform mean weights (ablation) | batch-wise loss · ModelNet40 | mAP: 83.3% -> 83.8 % | Helped |
| add optimal transport-derived importance weights rather than uniform mean weights (ablation) | batch-wise loss · ModelNet40 | accuracy: 89.7% -> 90.3 % | Helped |
| add optimal transport-derived importance weights rather than random weights (ablation) | batch-wise loss · ModelNet40 | mAP: 83.1% -> 83.8 % | Helped |
| add optimal transport-derived importance weights rather than random weights (ablation) | batch-wise loss · ModelNet40 | accuracy: 89.0% -> 90.3 % | Helped |
| replace pair-wise training with random pair sampling (ablation) | metric learning · ModelNet40 | mAP: 74.1% -> 83.1% | Helped |
| replace individual pair loss with triplet loss (ablation) | pair-wise metric learning · ModelNet40 | mAP: 69.3% -> 74.1% | Helped |
| use batch-wise optimal transport loss rather than pair-wise contrastive loss | 2D image recognition · CIFAR10 | mAP: approximately 15 % | Helped |
Closest papers by meaning
- Improved Deep Metric Learning with Multi-class N-pair Loss Objective (NeurIPS 2016 · similarity 0.87)
- Triplet-Center Loss for Multi-View 3D Object Retrieval (CVPR 2018 · similarity 0.87)
- Deep Correlated Metric Learning for Sketch-based 3D Shape Retrieval (AAAI 2017 · similarity 0.86)
- Divide and Conquer the Embedding Space for Metric Learning (CVPR 2019 · similarity 0.86)
- Deep Metric Learning Beyond Binary Supervision (CVPR 2019 · similarity 0.86)
- Working hard to know your neighbor's margins: Local descriptor learning loss (NeurIPS 2017 · similarity 0.86)
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.