Research map / Video / Video segmentation and tracking
DynamicStereo: Consistent Dynamic Depth From Stereo Videos
Nikita Karaev, Ignacio Rocco, Benjamin Graham, Natalia Neverova, Andrea Vedaldi, Christian Rupprecht
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 |
|---|---|---|---|
| process stereo videos jointly with DynamicStereo rather than estimating each frame independently | DynamicStereo · Sintel Stereo Clean | TEPE: 0.92 -> 0.77 | Improved |
| process stereo videos jointly with DynamicStereo rather than estimating each frame independently | DynamicStereo · Sintel Stereo Final | TEPE: 2.10 -> 1.45 | Improved |
| process stereo videos jointly with DynamicStereo rather than estimating each frame independently | DynamicStereo · Dynamic Replica, first 150 frames | TEPE: 0.145 -> 0.087 | Improved |
| train on Dynamic Replica instead of SceneFlow alone | stereo methods evaluated for temporal consistency · Dynamic Replica | TEPE | Improved |
| use separate update blocks for each resolution instead of sharing weights (ablation) | DynamicStereo · Dynamic Replica | δ1px: 6.48 -> 4.76 | Improved |
| use separate update blocks for each resolution instead of sharing weights (ablation) | DynamicStereo · Sintel Clean | δ3px: 6.60 -> 6.24 | Improved |
| use 3D convolution across space and time instead of 2D convolution in the update block (ablation) | DynamicStereo · Sintel Clean | TEPE: 1.05 -> 0.823 | Improved |
| use 3D convolution across space and time instead of 2D convolution in the update block (ablation) | DynamicStereo · Dynamic Replica | δ1px: 7.31 -> 4.76 | Improved |
| add space, stereo, and time attention in the SST block rather than using no attention (ablation) | DynamicStereo · Dynamic Replica | Bad1px: 7.31 -> 5.50 | Improved |
| add space, stereo, and time attention in the SST block rather than using no attention (ablation) | DynamicStereo · Sintel Clean | TEPE: 0.779 -> 0.753 | Improved |
| add space and time attention to the update block instead of using no attention (ablation) | DynamicStereo · Dynamic Replica | Bad1px: 7.41 -> 5.50 | Improved |
| add space and time attention to the update block instead of using no attention (ablation) | DynamicStereo · Sintel Clean | TEPE: 0.940 -> 0.753 | Improved |
Later papers that compared with this method
3 later accepted papers report DynamicStereo as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- Match Stereo Videos via Bidirectional Alignment (TPAMI 2026 · better on 7 of 7 shared results)
- Stereo Any Video: Temporally Consistent Stereo Matching (ICCV 2025 · better on 7 of 7 shared results)
- Match-Stereo-Videos: Bidirectional Alignment for Consistent Dynamic Stereo Matching (ECCV 2024 · better on 5 of 5 shared results)
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
- PPMStereo: Pick-and-Play Memory Construction for Consistent Dynamic Stereo Matching (NeurIPS 2025 · similarity 0.91)
- Stereo4D: Learning How Things Move in 3D from Internet Stereo Videos (CVPR 2025 · similarity 0.89)
- Align3R: Aligned Monocular Depth Estimation for Dynamic Videos (CVPR 2025 · similarity 0.88)
- 4D3R: Motion-Aware Neural Reconstruction and Rendering of Dynamic Scenes from Monocular Videos (NeurIPS 2025 · similarity 0.88)
- SpatialDreamer: Self-supervised Stereo Video Synthesis from Monocular Input (CVPR 2025 · similarity 0.87)
- World from Motion: Generative Dynamic Gaussian Reconstruction from Monocular Video (NeurIPS 2026 · 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.
