Research map / Video / Video segmentation and tracking
Stereo Any Video: Temporally Consistent Stereo Matching
Junpeng Jing, Weixun Luo, Ye Mao, Krystian Mikolajczyk
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 Video Depth Anything (VDA-S) foundation priors (ablation) | Stereo Any Video · Sintel Final | TEPE: 1.47 -> 1.27 | Improved |
| use VDA-S instead of image-based DepthAnythingV2-S (ablation) | Stereo Any Video · Sintel Final | TEPE: 1.33 -> 1.27 | Improved |
| use larger VDA-L backbone instead of VDA-S (ablation) | Stereo Any Video · Sintel Final | delta_1px^t: 10.98 -> 12.47 | Hurt |
| use 1D + 2D all-to-all-pairs correlation instead of local correlation (ablation) | Stereo Any Video · Sintel Final | TEPE: 1.27 -> 1.21 | Improved |
| extend all-to-all-pairs correlation to 3D search window (ablation) | Stereo Any Video · Sintel Final | TEPE: 1.21 -> 4.51 | Hurt |
| add MLP before convolution in cost volume encoder (ablation) | Stereo Any Video · Sintel Final | TEPE: 1.26 -> 1.21 | Improved |
| use temporal convex upsampling instead of bilinear (ablation) | Stereo Any Video · Sintel Final | TEPE: 1.26 -> 1.04 | Improved |
| use temporal convex upsampling instead of spatial convex (ablation) | Stereo Any Video · Sintel Final | TEPE: 1.21 -> 1.04 | Improved |
| add temporal and spatial attention in GRU (ablation) | Stereo Any Video · Dynamic Replica | TEPE: 0.067 -> 0.057 | Improved |
| add temporal and spatial attention in GRU (ablation) | Stereo Any Video · Sintel Final | TEPE: 1.04 -> 1.07 | Hurt |
| add extra 1/32 starting resolution at inference (ablation) | Stereo Any Video · Sintel Final | delta_1px^t: 9.98 -> 10.82 | Hurt |
| use Stereo Any Video instead of BiDAStereo, zero-shot | video stereo matching · Sintel Final | TEPE: 1.26 -> 1.07 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DynamicStereo — DynamicStereo: Consistent Dynamic Depth From Stereo Videos (CVPR 2023) · better on 7 of 7 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
- Temporally Consistent Stereo Matching (ECCV 2024 · similarity 0.92)
- Match-Stereo-Videos: Bidirectional Alignment for Consistent Dynamic Stereo Matching (ECCV 2024 · similarity 0.92)
- Any-Stereo: Arbitrary Scale Disparity Estimation for Iterative Stereo Matching (AAAI 2024 · similarity 0.92)
- Stereo Anywhere: Robust Zero-Shot Deep Stereo Matching Even Where Either Stereo or Mono Fail (CVPR 2025 · similarity 0.90)
- PPMStereo: Pick-and-Play Memory Construction for Consistent Dynamic Stereo Matching (NeurIPS 2025 · similarity 0.89)
- Video Depth Anything: Consistent Depth Estimation for Super-Long Videos (CVPR 2025 · 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.
