Research map / 3D vision / Depth and stereo
LinStereo: Linear-Complexity Global Attention for Multi-Scale Iterative Stereo Matching
Yiran Wang, Oliver Turner, Viorela Ila
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 ConvGRU with global linear attention plus positional encodings and gating (full PALA vs plain linear attention) (ablation) | LinStereo · KITTI 2015 | EPE: 1.18 -> 1.01 | Improved |
| use asymmetric rather than symmetric RoPE (ablation) | PALA · KITTI 2015 | EPE: 1.05 -> 1.01 | Improved |
| add PALA to HSCV+DPI (ablation) | LinStereo · KITTI 2015 | EPE: 1.21 -> 1.01 | Improved |
| use per-scale hierarchical cost volumes instead of pooled multi-scale feature (ablation) | LinStereo · KITTI 2015 | EPE: 1.07 -> 1.01 | Improved |
| add Depth Prior Initialization (ablation) | LinStereo · KITTI 2015 | EPE: 1.10 -> 1.01 | Improved |
| raise iterations from 8 to 12 (ablation) | LinStereo · KITTI 2015 | EPE: 1.01 vs 1.03 | No clear effect |
| reduce iterations from 8 to 2 (ablation) | LinStereo · KITTI 2015 | EPE: 1.01 -> 1.24 | Hurt |
| use 3 update scales instead of 1 (ablation) | LinStereo · KITTI 2015 | EPE: 1.13 -> 1.01 | Improved |
| stack more PALA blocks per updater (ablation) | PALA · KITTI 2015 | EPE: 1.01 -> 1.05 | Hurt |
| learning rate 4e-4 instead of 2e-4 (ablation) | LinStereo · KITTI 2015 | EPE: 1.01 -> 1.16 | Hurt |
| PALA vs ConvGRU update operator latency | LinStereo · per-iteration latency | ms: 3.50 vs 3.63 and 3.43 | No clear effect |
| global linear-attention decoder on VFM backbone (SceneFlow only) vs larger extra-data methods | LinStereo · TartanAir-UW | AbsRel: 0.05 -> 0.04 | 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.
- DEFOM-Stereo — DEFOM-Stereo: Depth Foundation Model Based Stereo Matching (CVPR 2025) · better on 1 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
- AquaStereo: Enabling Underwater Stereo Matching via Depth-Conditioned Diffusion and Geometry Self-Distillation (ECCV 2026 · similarity 0.92)
- NAUTILUS: A Large Multimodal Model for Underwater Scene Understanding (NeurIPS 2025 · similarity 0.87)
- Atlantis: Enabling Underwater Depth Estimation with Stable Diffusion (CVPR 2024 · similarity 0.86)
- Any-Stereo: Arbitrary Scale Disparity Estimation for Iterative Stereo Matching (AAAI 2024 · similarity 0.85)
- DepthMaster: Unified Monocular Depth Estimation for Perspective and Panoramic Images (NeurIPS 2026 · similarity 0.85)
- 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.
