Research map / 3D vision / Depth and stereo
Diving into the Fusion of Monocular Priors for Generalized Stereo Matching
Chengtang Yao, Lidong Yu, Zhidan Liu, Jiaxi Zeng, Yuwei Wu, Yunde Jia
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 DepthAnything v2 monocular encoder (context features) to RAFT-Stereo (ablation) | RAFT-Stereo · Middlebury (H) | bad 2.0: 14.12 -> 9.81 | Improved |
| keep RAFT-Stereo context features (vs removing them) (ablation) | RAFT-Stereo · Middlebury (H) | bad 2.0: 12.45 -> 14.12 | Hurt |
| add iterative local fusion (ILF) to Baseline + ME (ablation) | Baseline + ME · Middlebury (H) | bad 2.0: 9.81 -> 9.06 | Improved |
| replace iterative direct fusion with local ordering fusion (ablation) | Baseline + ME · Middlebury (H) | bad 2.0: 10.34 -> 9.06 | Improved |
| add iterative direct fusion (IDF) (ablation) | Baseline + ME · Middlebury (H) | bad 2.0: 9.81 -> 10.34 | Hurt |
| add global fusion (GF) with registration (ablation) | Baseline + ME + ILF · Middlebury (H) | bad 2.0: 9.06 -> 8.35 | Improved |
| use global fusion instead of post-fusion without registration (ablation) | Baseline + ME · Middlebury (H) | bad 2.0: 9.71 -> 8.35 | Improved |
| add registration of monocular depth in global fusion (ablation) | Global fusion · Middlebury (H) | bad 2.0: 9.69 -> 8.77 | Improved |
| use hybrid confidence (cost, hidden state, guidance) instead of cost-only (ablation) | Global fusion · Middlebury (H) | bad 2.0: 8.77 -> 8.35 | Improved |
| use fixed-weight LBP-like convolution instead of learnable convolution (ablation) | Iterative local fusion · Middlebury (H) | bad 2.0: 9.53 -> 9.06 | Improved |
| use LBP kernel sizes 5,3 instead of 9,7,5,3 (ablation) | Iterative local fusion · Middlebury (H) | bad 2.0: 9.57 -> 9.06 | Improved |
| vary amplitude parameter r from 1 to 3 (ablation) | Iterative local fusion · Middlebury (H) | bad 2.0: 9.06 -> 9.42 | No clear effect |
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.
- Mocha-Stereo — MoCha-Stereo: Motif Channel Attention Network for Stereo Matching (CVPR 2024) · 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
- Single View Stereo Matching (CVPR 2018 · similarity 0.92)
- S2M2: Scalable Stereo Matching Model for Reliable Depth Estimation (ICCV 2025 · similarity 0.90)
- Stereo Anywhere: Robust Zero-Shot Deep Stereo Matching Even Where Either Stereo or Mono Fail (CVPR 2025 · similarity 0.90)
- MonSter: Marry Monodepth to Stereo Unleashes Power (CVPR 2025 · similarity 0.89)
- Any-Stereo: Arbitrary Scale Disparity Estimation for Iterative Stereo Matching (AAAI 2024 · similarity 0.88)
- Stereo Risk: A Continuous Modeling Approach to Stereo Matching (ICML 2024 · 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.
