Research map / 3D vision / Depth and stereo
Learning Two-View Correspondences and Geometry Using Order-Aware Network
Jiahui Zhang, Dawei Sun, Zixin Luo, Anbang Yao, Lei Zhou, Tianwei Shen, Yurong Chen, Long Quan, Hongen Liao
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 Order-Aware DiffPool/DiffUnpool layers (UnB) (ablation) | PointCN · YFCC100M outdoor unknown, no RANSAC | mAP5: 23.55 -> 28.78 | Improved |
| add plain DiffUnpool layers (UnA) (ablation) | PointCN · YFCC100M outdoor unknown, no RANSAC | mAP5: 23.55 -> 24.10 | No clear effect |
| use Order-Aware DiffUnpool instead of plain DiffUnpool (ablation) | PointCN + pooling · YFCC100M outdoor unknown, no RANSAC | mAP5: 24.10 -> 28.78 | Improved |
| replace PointCN blocks at second level with Order-Aware Filtering blocks (ablation) | PointCN + UnB · YFCC100M outdoor unknown, no RANSAC | mAP5: 28.78 -> 32.55 | Improved |
| use larger three-level model (ablation) | Order-Aware Network · YFCC100M outdoor unknown, no RANSAC | mAP5: 32.55 -> 30.48 | Hurt |
| use geometry loss instead of L2 loss (ablation) | Order-Aware Network · YFCC100M outdoor unknown | mAP5: 51.63 -> 51.10 with RANSAC; 32.55 -> 33.68 without | Mixed |
| add iterative network (ablation) | Order-Aware Network + geometry loss · YFCC100M outdoor unknown, no RANSAC | mAP5: 33.68 -> 39.33 | Improved |
| lower RANSAC inlier threshold from 0.01 to 0.001 (ablation) | PointCN · YFCC100M outdoor unknown, with RANSAC | mAP5: 17.53 -> 44.50 | Improved |
| train on 68 sequences instead of 2 (ablation) | PointCN · YFCC100M outdoor unknown, with RANSAC | mAP5: 44.50 -> 47.98 | Improved |
| use SuperPoint instead of SIFT features (ablation) | PointCN · YFCC100M outdoor unknown, with RANSAC | mAP5: 47.98 -> 43.18 | Hurt |
| Order-Aware Network versus PointCN | PointCN · YFCC100M outdoor unknown, no RANSAC | mAP5: 23.55 -> 32.55 | Improved |
Closest papers by meaning
- OANet: Learning Two-View Correspondences and Geometry Using Order-Aware Network (TPAMI 2020 · similarity 0.97)
- T-Net: Effective Permutation-Equivariant Network for Two-View Correspondence Learning (ICCV 2021 · similarity 0.90)
- U-Match: Exploring Hierarchy-Aware Local Context for Two-View Correspondence Learning (TPAMI 2024 · similarity 0.90)
- Scalable and Generalizable Correspondence Pruning via Geometry-Consistent Pre-Training (TPAMI 2026 · similarity 0.88)
- Two-View Geometry Scoring Without Correspondences (CVPR 2023 · similarity 0.88)
- BCLNet: Bilateral Consensus Learning for Two-View Correspondence Pruning (AAAI 2024 · 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.
