Research map / Architectures and efficiency / RNNs and normalization
Bilateral Grid Learning for Stereo Matching Networks
Bin Xu, Yuhua Xu, Xiaoli Yang, Wei Jia, Yulan Guo
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 linear cost-volume upsampling with CUBG (ablation) | BGNet · SceneFlow Finalpass | EPE: 1.40 -> 1.17 | Improved |
| Replace linear cost-volume upsampling with CUBG (ablation) | BGNet · SceneFlow Finalpass, edge regions | EPE-edge: 8.13 -> 5.95 | Improved |
| Replace linear cost-volume upsampling with CUBG (ablation) | BGNet · SceneFlow Finalpass, flat regions | EPE-flat: 0.71 -> 0.68 | No clear effect |
| Replace the learned guidance map with input-image luma (ablation) | BGNet CUBG · SceneFlow | EPE: 1.17 -> 1.28 | Hurt |
| Build the cost volume at 1/16 rather than the original resolution (ablation) | BGNet · SceneFlow | EPE: 1.17 -> 1.58 | Hurt |
| Build the cost volume at 1/16 rather than the original resolution (ablation) | BGNet · SceneFlow | runtime (ms): 25.3 ms to 17.1 ms | Improved |
| Embed CUBG cost-volume upsampling | GCNet · SceneFlow | EPE: 2.51 -> 1.07 | Improved |
| Embed CUBG cost-volume upsampling | PSMNet · SceneFlow | EPE: 1.09 -> 0.92 | Improved |
| Embed CUBG cost-volume upsampling | GANet_deep · SceneFlow | EPE: 0.95 -> 0.63 | Improved |
| Embed CUBG and replace the narrow-volume pipeline with full-volume aggregation | DeepPrunerFast · SceneFlow | EPE: 0.97 -> 0.84 | Improved |
| Add an hourglass disparity refinement module (ablation) | BGNet · KITTI 2015 test | D1-all: 2.51 -> 2.19 | Improved |
| Train with IRS in addition to Flyingthings3D | BGNet and BGNet+ · Middlebury 2014 | Bad 2.0 (%): 17.5 -> 13.5; 17.2 -> 11.6 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- GANet — GA-Net: Guided Aggregation Net for End-To-End Stereo Matching (CVPR 2019) · better on 0 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- StereoNet: Guided Hierarchical Refinement for Real-Time Edge-Aware Depth Prediction (ECCV 2018 · similarity 0.89)
- CFNet: Cascade and Fused Cost Volume for Robust Stereo Matching (CVPR 2021 · similarity 0.88)
- AANet: Adaptive Aggregation Network for Efficient Stereo Matching (CVPR 2020 · similarity 0.88)
- HITNet: Hierarchical Iterative Tile Refinement Network for Real-time Stereo Matching (CVPR 2021 · similarity 0.88)
- Stereo Matching Using Multi-Level Cost Volume and Multi-Scale Feature Constancy (TPAMI 2019 · similarity 0.87)
- IINet: Implicit Intra-inter Information Fusion for Real-Time Stereo Matching (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.
