Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
AquaStereo: Enabling Underwater Stereo Matching via Depth-Conditioned Diffusion and Geometry Self-Distillation
Qizhe Wei, Yingping Liang, Shao You, Ying Fu
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 |
|---|---|---|---|
| train on diffusion-rendered UW-Dataset instead of SceneFlow | IGEV++ · FLSea-Canyon | EPE: 5.41 -> 3.05 | Improved |
| train on UW-Dataset instead of StereoAdapter CG rendering | IGEV++ · Squid | EPE: 1.86 -> 1.80 | Improved |
| train on UW-Dataset instead of StereoAdapter | IGEV++ · TartanAir (underwater) | EPE: 0.52 -> 0.55 | Hurt |
| add terrestrial KITTI to SceneFlow | IGEV++ · Squid | EPE: 2.18 -> 2.02 | No clear effect |
| enable left-right consistency module in generation (ablation) | stereo data generation · UWStereo | EPE: 3.237 -> 2.953 | Improved |
| enable LR-consistency module (ablation) | stereo data generation · UWStereo | Break px: 15 -> 5 | Improved |
| add cross-domain self-distillation (parameter-disjoint) (ablation) | IGEV++ · UWStereo Total | EPE: 2.9530 -> 2.7510 | Improved |
| add clean-branch supervision with shared pseudo target (ablation) | self-distillation · UWStereo Total | EPE: 2.7510 -> 2.6350 | Improved |
| add L2 regularization to self-distillation (ablation) | self-distillation · UWStereo Total | EPE: 2.7510 -> 3.0510 | Hurt |
| share extractor, refinement or aggregation between teacher and student (ablation) | self-distillation · UWStereo Total | EPE: 2.7510 -> 2.7200 | No clear effect |
| replace IGEV++ backbone with perception frames + video encoder + DINOv2 (ablation) | IGEV++ · UWStereo Total | EPE: 2.9530 -> 2.4830 | Improved |
| replace encoder with VGG19+DINOv2 (without perception frames) (ablation) | IGEV++ · UWStereo Total | EPE: 2.9530 -> 3.0790 | Hurt |
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
- LinStereo: Linear-Complexity Global Attention for Multi-Scale Iterative Stereo Matching (ECCV 2026 · similarity 0.92)
- Atlantis++: Enabling Underwater Depth Estimation with Stable Diffusion and Beyond (IJCV 2026 · similarity 0.88)
- Atlantis: Enabling Underwater Depth Estimation with Stable Diffusion (CVPR 2024 · similarity 0.88)
- NAUTILUS: A Large Multimodal Model for Underwater Scene Understanding (NeurIPS 2025 · similarity 0.86)
- Plenodium: Underwater 3D Scene Reconstruction with Plenoptic Medium Representation (NeurIPS 2025 · similarity 0.86)
- WaterDiffusion: Learning a Prior-involved Unrolling Diffusion for Joint Underwater Saliency Detection and Visual Restoration (AAAI 2025 · similarity 0.85)
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.
