Research map / 3D vision / Depth and stereo
StereoNet: Guided Hierarchical Refinement for Real-Time Edge-Aware Depth Prediction
Sameh Khamis, Sean Fanello, Christoph Rhemann, Adarsh Kowdle, Julien Valentin, Shahram Izadi
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 hierarchical multi-level refinement instead of a single refinement (8x) (ablation) | StereoNet 8x · Scene Flow | EPE all: 1.532 -> 1.101 | Improved |
| add hierarchical multi-level refinement instead of a single refinement (16x) (ablation) | StereoNet 16x · Scene Flow | EPE all: 1.974 -> 1.525 | Improved |
| use finer cost volume (8x instead of 16x) with multi refinement (ablation) | StereoNet multi · Scene Flow | EPE all: 1.525 -> 1.101 | Improved |
| use finer cost volume (8x instead of 16x) with single refinement (ablation) | StereoNet single · Scene Flow | EPE all: 1.974 -> 1.532 | Improved |
| add refinement to the 8x multi cost-volume output (ablation) | StereoNet 8x multi · Scene Flow | EPE all: 2.512 -> 1.101 | Improved |
| add refinement to the 8x single cost-volume output (ablation) | StereoNet 8x single · Scene Flow | EPE all: 2.486 -> 1.532 | Improved |
| add refinement to the 16x multi cost-volume output (ablation) | StereoNet 16x multi · Scene Flow | EPE all: 3.764 -> 1.525 | Improved |
| add refinement to the 16x single cost-volume output (ablation) | StereoNet 16x single · Scene Flow | EPE all: 3.558 -> 1.974 | Improved |
| use soft argmin instead of sampling-based probabilistic selection (ablation) | StereoNet · Scene Flow | convergence and accuracy | Improved |
| use unrefined 1/8 cost volume vs the prior full model | Prior method · Scene Flow | EPE all: 2.51 -> 2.48 | Improved |
| use multi-scale 8x model vs CRL | CRL · Scene Flow | EPE all: 1.32 -> 1.1 | Improved |
| StereoNet vs state of the art on KITTI 2012 | Prior state of the art · KITTI 2012 | Out-Noc (%): 2.71 -> 4.91 | Hurt |
Later papers that compared with this method
5 later accepted papers report StereoNet as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- BANet: Bilateral Aggregation Network for Mobile Stereo Matching (ICCV 2025 · better on 1 of 1 shared result)
- Facial Depth and Normal Estimation Using Single Dual-Pixel Camera (ECCV 2022 · better on 4 of 4 shared results)
- Displacement-Invariant Cost Computation for Stereo Matching (IJCV 2022 · better on 2 of 2 shared results)
- EDNet: Efficient Disparity Estimation With Cost Volume Combination and Attention-Based Spatial Residual (CVPR 2021 · better on 0 of 1 shared result)
- HITNet: Hierarchical Iterative Tile Refinement Network for Real-time Stereo Matching (CVPR 2021 · 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
- AANet: Adaptive Aggregation Network for Efficient Stereo Matching (CVPR 2020 · similarity 0.90)
- Du²Net: Learning Depth Estimation from Dual-Cameras and Dual-Pixels (ECCV 2020 · similarity 0.90)
- On the confidence of stereo matching in a deep-learning era: a quantitative evaluation (TPAMI 2021 · similarity 0.90)
- A Survey on Deep Learning Techniques for Stereo-Based Depth Estimation (TPAMI 2020 · similarity 0.90)
- EdgeStereo: An Effective Multi-task Learning Network for Stereo Matching and Edge Detection (IJCV 2020 · similarity 0.90)
- Stereo Risk: A Continuous Modeling Approach to Stereo Matching (ICML 2024 · similarity 0.89)
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.
