Research map / Architectures and efficiency / CNN design
Improved Stereo Matching With Constant Highway Networks and Reflective Confidence Learning
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 |
|---|---|---|---|
| replace MC-CNN with the full stereo pipeline | Stereo matching · KITTI 2012 test (NOC) | error rate: 2.43 -> 2.29 | Helped |
| replace MC-CNN with the full stereo pipeline | Stereo matching · KITTI 2015 test (ALL) | error rate: 3.89 -> 3.42 | Helped |
| use the fast stereo pipeline instead of MC-CNN-fast | Stereo matching · KITTI 2015 under 5s (ALL) | error rate: 4.62 -> 3.78 | Helped |
| replace MC-CNN matching architecture with λ-ResMatch | Stereo matching · KITTI 2012 validation | error rate: 2.84 -> 2.71 | Helped |
| replace MC-CNN matching architecture with λ-ResMatch | Stereo matching · KITTI 2015 validation | error rate: 3.53 -> 3.35 | Helped |
| add the global disparity network in place of winner-takes-all (ablation) | λ-ResMatch pipeline · KITTI 2015 validation (accurate) | error rate: 3.15 -> 2.87 | Helped |
| add confidence-guided refinement (ablation) | λ-ResMatch + global disparity network · KITTI 2015 validation (accurate) | error rate: 2.87 -> 2.83 | Helped |
| add confidence-guided refinement (ablation) | λ-ResMatch + global disparity network · KITTI 2012 validation (accurate) | error rate: 2.40 -> 2.38 | Helped |
| add hybrid loss to the matching network (ablation) | λ-ResMatch · KITTI 2012 validation | error rate: 2.73 -> 2.71 | Helped |
| add hybrid loss to the matching network (ablation) | λ-ResMatch · KITTI 2015 validation | error rate: 3.42 -> 3.35 | Helped |
| replace identity shortcuts with constant λ shortcuts in the inner blocks (ablation) | Our multilevel residual architecture · KITTI 2012 validation | error rate: 2.78 -> 2.75 | Helped |
| replace identity shortcuts with constant λ shortcuts in the outer blocks (ablation) | Our multilevel residual architecture · KITTI 2015 validation | error rate: 3.49 -> 3.46 | Helped |
Closest papers by meaning
- AANet: Adaptive Aggregation Network for Efficient Stereo Matching (CVPR 2020 · similarity 0.90)
- Local Similarity Pattern and Cost Self-Reassembling for Deep Stereo Matching Networks (AAAI 2022 · similarity 0.90)
- Modeling Stereo-Confidence out of the End-to-End Stereo-Matching Network via Disparity Plane Sweep (AAAI 2024 · similarity 0.90)
- Stereo Risk: A Continuous Modeling Approach to Stereo Matching (ICML 2024 · similarity 0.89)
- Hierarchical Object-Aware Dual-Level Contrastive Learning for Domain Generalized Stereo Matching (NeurIPS 2024 · similarity 0.88)
- Robust Synthetic-to-Real Transfer for Stereo Matching (CVPR 2024 · similarity 0.88)
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.