Research map / 3D vision / Depth and stereo
On the Synergies between Machine Learning and Binocular Stereo for Depth Estimation from Images: a Survey
Matteo Poggi, Fabio Tosi, Konstantinos Batsos, Philippos Mordohai, Stefano Mattoccia
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 learned matching cost (MC-CNN) over classical SGBM | stereo pipeline · KITTI 2015 leaderboard | D1-all error (%): 10.86 -> 3.89 | Improved |
| build learned SGM-Net optimization on MC-CNN | MC-CNN-acrt · KITTI 2015 leaderboard | D1-all error (%): 3.89 -> 3.66 | Improved |
| add learned disparity refinement on top of other algorithms | stereo pipeline · KITTI 2015 leaderboard | D1-all error (%): 3.89 -> 3.03 | Improved |
| use end-to-end CSPN network instead of MC-CNN pipeline | MC-CNN-acrt · KITTI 2015 leaderboard | D1-all error (%): 3.89 -> 1.74 | Improved |
| use real-time end-to-end networks (DispNet-C, MADNet, StereoNet) | end-to-end stereo · KITTI 2015 leaderboard | D1-all error (%): 4.34, 4.66, 4.83 vs 3.89 | Hurt |
| use 3D cost-volume architectures instead of 2D | end-to-end stereo · KITTI 2015 leaderboard | D1-all error (%): 1.74 vs 2.02 | Mixed |
| use end-to-end HSM instead of MC-CNN on high-res images | MC-CNN-acrt · Middlebury 2014 leaderboard | bad 2.0 error (%): 8.08 -> 10.2 | Hurt |
| use end-to-end HSM instead of MC-CNN on high-res images | MC-CNN-acrt · Middlebury 2014 leaderboard | average pixel error: 3.82 -> 2.07 | Improved |
| use end-to-end DispNet-CSS instead of SGM | SGM · ETH3D leaderboard | bad 1.0 error (%): 10.08 -> 2.69 | Improved |
| use end-to-end DispNet-CSS instead of SGM | SGM · Middlebury 2014 leaderboard | bad 2.0 error (%): 18.4 -> 22.8 | Hurt |
| use larger receptive field confidence network LGC-Net instead of O1 | confidence estimation · KITTI 2015 confidence (SGM disparities) | AUC x10^2: 4.61 -> 3.92 | Improved |
| add cost volume cues to image and disparity cues (LAF-Net) | LGC-Net · KITTI 2015 confidence (SGM disparities) | AUC x10^2: 3.92 -> 3.85 | Improved |
Closest papers by meaning
- A Survey on Deep Learning Techniques for Stereo-Based Depth Estimation (TPAMI 2020 · similarity 0.94)
- Stereo Matching by Training a Convolutional Neural Network to Compare Image Patches (JMLR 2016 · similarity 0.92)
- On the confidence of stereo matching in a deep-learning era: a quantitative evaluation (TPAMI 2021 · similarity 0.92)
- Du²Net: Learning Depth Estimation from Dual-Cameras and Dual-Pixels (ECCV 2020 · similarity 0.91)
- A Survey on Deep Stereo Matching in the Twenties (IJCV 2025 · similarity 0.91)
- Stereo Risk: A Continuous Modeling Approach to Stereo Matching (ICML 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.
