Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
PromptStereo: Zero-Shot Stereo Matching via Structure and Motion Prompts
Xianqi Wang, Hao Yang, Hangtian Wang, Junda Cheng, Gangwei Xu, Min Lin, Xin Yang
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 |
|---|---|---|---|
| replace dual-branch GRU with Prompt Recurrent Unit plus motion prompt (ablation) | MonSter · Midd-T Bad 2.0 | Bad 2.0 (%): 7.27 -> 4.18 | Improved |
| add Structure Prompt (ablation) | PRU + MP · Midd-T Bad 2.0 | Bad 2.0 (%): 4.18 -> 3.90 | Improved |
| add Affine-Invariant Fusion of initial disparity and relative depth (ablation) | PRU + MP + SP · Midd-T Bad 2.0 | Bad 2.0 (%): 3.90 -> 3.76 | Improved |
| replace GRU with PRU in RAFT-Stereo (ablation) | RAFT-Stereo · KITTI 2015 Bad 3.0 | Bad 3.0 (%): 5.68 -> 4.78 | Improved |
| replace GRU with PRU in IGEV-Stereo (ablation) | IGEV-Stereo · KITTI 2015 Bad 3.0 | Bad 3.0 (%): 6.03 -> 4.84 | Improved |
| remove pre-trained DPT decoder weights from PRU (ablation) | PromptStereo · Midd-T Bad 2.0 | Bad 2.0 (%): 4.39 -> 3.76 | Improved |
| merge hidden state and prompt into one conv block (ablation) | PromptStereo · Midd-T Bad 2.0 | Bad 2.0 (%): 3.76 -> 4.37 | Hurt |
| use zero-initialized last conv in prompt block (ablation) | PromptStereo · Midd-T Bad 2.0 | Bad 2.0 (%): 3.76 -> 4.04 | Hurt |
| use PRU instead of GRU at 4 iterations (ablation) | PromptStereo vs MonSter · Midd-2021 Bad 2.0 | Bad 2.0 (%): 10.75 -> 4.35 | Improved |
| increase inference iterations from 16 to 32 (ablation) | PromptStereo · Midd-2021 Bad 2.0 | Bad 2.0 (%): 2.79 -> 2.78 | No clear effect |
| increase inference iterations from 16 to 32 (ablation) | PromptStereo · Booster (Q) Bad 2.0 | Bad 2.0 (%): 4.09 -> 3.67 | Improved |
| train on unlimited mixed datasets instead of Scene Flow only | PromptStereo · Booster (Q) Bad 2.0 | Bad 2.0 (%): 12.13 -> 3.67 | 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.
- FoundationStereo — FoundationStereo: Zero-Shot Stereo Matching (CVPR 2025) · better on 2 of 2 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
- FoundationStereo: Zero-Shot Stereo Matching (CVPR 2025 · similarity 0.91)
- Stereo Anywhere: Robust Zero-Shot Deep Stereo Matching Even Where Either Stereo or Mono Fail (CVPR 2025 · similarity 0.91)
- Zero-Shot Event-Intensity Asymmetric Stereo via Visual Prompting from Image Domain (NeurIPS 2024 · similarity 0.89)
- Any-Stereo: Arbitrary Scale Disparity Estimation for Iterative Stereo Matching (AAAI 2024 · similarity 0.88)
- MonSter: Marry Monodepth to Stereo Unleashes Power (CVPR 2025 · similarity 0.88)
- RefineAny3D: Depth Refinement as Semantic Alignment for Monocular 3D Detection (NeurIPS 2026 · similarity 0.88)
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.
