Research map / 3D vision / Depth and stereo
Stereo Matching by Training a Convolutional Neural Network to Compare Image Patches
Jure Žbontar, Yann LeCun
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 all post-processing steps of the stereo method (ablation) | accurate architecture · KITTI 2012 (validation) | error (%): 13.49 -> 2.61 | Improved |
| add all post-processing steps of the stereo method (ablation) | accurate architecture · Middlebury (validation) | error (%): 28.33 -> 7.91 | Improved |
| add semiglobal matching (ablation) | accurate architecture · KITTI 2012 (validation) | error (%): 4.26 -> 2.61 | Improved |
| add semiglobal matching (ablation) | accurate architecture · Middlebury (validation) | error (%): 11.99 -> 7.91 | Improved |
| add data set augmentation (ablation) | accurate architecture · KITTI 2012 (validation) | error (%): 2.73 -> 2.61 | Improved |
| add data set augmentation (ablation) | accurate architecture · Middlebury (validation) | error (%): 8.75 -> 7.91 | Improved |
| add left-right consistency check (ablation) | accurate architecture · Middlebury (validation) | error (%): 7.91 -> 8.22 | Hurt |
| replace census transform matching cost with CNN (accurate) | stereo method · KITTI 2012 (validation) | error (%): 4.90 -> 2.61 | Improved |
| replace census transform matching cost with CNN (accurate) | stereo method · Middlebury (validation) | error (%): 16.72 -> 7.91 | Improved |
| increase training set size from 20% to 100% (ablation) | accurate architecture · KITTI 2012 (validation) | error (%): 2.84 -> 2.61 | Improved |
| increase training set size from 20% to 100% (ablation) | accurate architecture · Middlebury (validation) | error (%): 9.73 -> 7.91 | Improved |
| increase conv layers from 1 to 4 (ablation) | accurate architecture · KITTI 2012 (validation) | error (%): 3.97 -> 2.61 | Improved |
Closest papers by meaning
- Stereo Matching Using Multi-Level Cost Volume and Multi-Scale Feature Constancy (TPAMI 2019 · similarity 0.95)
- Unsupervised Learning of Stereo Matching (ICCV 2017 · similarity 0.95)
- Local Similarity Pattern and Cost Self-Reassembling for Deep Stereo Matching Networks (AAAI 2022 · similarity 0.93)
- Deep Stereo Matching With Explicit Cost Aggregation Sub-Architecture (AAAI 2018 · similarity 0.93)
- Learning for Disparity Estimation Through Feature Constancy (CVPR 2018 · similarity 0.92)
- Efficient Deep Learning for Stereo Matching (CVPR 2016 · similarity 0.92)
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.
