Research map / Label-efficient and robust learning / Domain adaptation and generalization
Domain-invariant Stereo Matching Networks
Feihu Zhang, Xiaojuan Qi, Ruigang Yang, Victor Prisacariu, Benjamin Wah, Philip Torr
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 |
|---|---|---|---|
| replace batch norm with domain normalization (ablation) | DSMNet · SceneFlow-trained, tested on KITTI | 2-pixel error (%): 9.4 -> 7.9 | Improved |
| replace batch norm with domain normalization (ablation) | DSMNet · SceneFlow-trained, tested on Middlebury | 3-pixel error (%): 30.3 -> 27.1 | Improved |
| add 3 non-local filter layers to features (ablation) | DSMNet with DN · SceneFlow-trained, tested on Middlebury | 3-pixel error (%): 27.1 -> 24.2 | Improved |
| increase feature non-local filter layers from 3 to 7 (ablation) | DSMNet with DN · SceneFlow-trained, tested on Middlebury | 3-pixel error (%): 24.2 -> 22.9 | Improved |
| increase feature non-local filter layers from 7 to 9 (ablation) | DSMNet with DN · SceneFlow-trained, tested on KITTI | 2-pixel error (%): 6.8 -> 6.8 | No clear effect |
| add 2 cost-volume non-local filter layers (ablation) | DSMNet with DN + 7 NLF · SceneFlow-trained, tested on Middlebury | 3-pixel error (%): 22.9 -> 21.8 | Improved |
| add domain normalization and non-local filters to PSMNet | PSMNet · SceneFlow-trained, tested on KITTI | 2-pixel error (%): 16.3 -> 8.5 | Improved |
| add domain normalization and non-local filters to GANet | GANet · SceneFlow-trained, tested on Middlebury | 3-pixel error (%): 32.2 -> 23.7 | Improved |
| use domain normalization instead of instance norm (ablation) | DSMNet · SceneFlow-trained, tested on KITTI | error rate (%): 6.4 -> 4.1 | Improved |
| use graph-based non-local filter instead of affinity propagation (ablation) | DSMNet · Middlebury full | error rate (%): 23.1 -> 20.1 | Improved |
| add non-local filtering layers to GANet-deep before KITTI fine-tuning | GANet-deep · KITTI 2015 benchmark (fine-tuned) | error rate (%), all area: 1.81 -> 1.77 | Improved |
| add Carla synthetic stereo data to SceneFlow training | DSMNet · KITTI 2012 | error rate (%): 6.2 -> 3.9 | Improved |
Closest papers by meaning
- Domain Generalized Stereo Matching via Hierarchical Visual Transformation (CVPR 2023 · similarity 0.92)
- Revisiting Domain Generalized Stereo Matching Networks From a Feature Consistency Perspective (CVPR 2022 · similarity 0.92)
- AdaStereo: An Efficient Domain-Adaptive Stereo Matching Approach (IJCV 2022 · similarity 0.92)
- GraftNet: Towards Domain Generalized Stereo Matching With a Broad-Spectrum and Task-Oriented Feature (CVPR 2022 · similarity 0.91)
- Hierarchical Object-Aware Dual-Level Contrastive Learning for Domain Generalized Stereo Matching (NeurIPS 2024 · similarity 0.91)
- Robust Synthetic-to-Real Transfer for Stereo Matching (CVPR 2024 · similarity 0.90)
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.
