Research map / 3D vision / Depth and stereo
MonSter++: Unified Stereo Matching, Multi-View Stereo, and Real-Time Stereo With Monodepth Priors
Junda Cheng, Wenjing Liao, Zhipeng Cai, Longliang Liu, Gangwei Xu, Xianqi Wang, Yuzhou Wang, Zikang Yuan, Yong Deng, Jinliang Zang, Yangyang Shi, Jinhui Tang and 1 more
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 Mono Guided Refinement instead of conv hourglass fusion (ablation) | MonSter++ · Scene Flow test | EPE: 6.52% | Improved |
| add Stereo Guided Alignment for per-pixel scale and shift on top of MGR (ablation) | MonSter++ · Scene Flow test | EPE: 9.30% in EPE and 10.69% in 1-pixel error | Improved |
| use SGA instead of conv for shift refinement (ablation) | MonSter++ · Scene Flow test | 1-pixel error: 8.09% improvement in 1-pixel error | Improved |
| add feature sharing of monocular ViT features with stereo branch (ablation) | MonSter++ · Scene Flow test | EPE: 0.39 -> 0.37 | Improved |
| add monocular branch with conv fusion to IGEV (ablation) | IGEV · Scene Flow test | EPE: 0.47 -> 0.46 | Improved |
| use DepthAnythingV2 instead of MiDaS as monocular branch (ablation) | MonSter++ · Scene Flow test | EPE: 0.41 -> 0.37 | Improved |
| use DepthAnythingV2 instead of DepthAnythingV1 as monocular branch (ablation) | MonSter++ · Scene Flow test | EPE: 0.39 -> 0.37 | Improved |
| reduce inference iterations from 32 to 4 (ablation) | MonSter++ · Scene Flow test | EPE: 0.37 -> 0.42 | Hurt |
| replace IGEV with MonSter++ using 4 iterations instead of 32 | MonSter++ vs IGEV · Scene Flow test | EPE and run-time: 0.47 -> 0.42; 0.37s -> 0.34s | Improved |
| add monocular prior fusion (MonSter++ over IGEV) | IGEV · Zero-shot avg of 4 datasets (Scene Flow) | error rate: 5.03 -> 3.70 | Improved |
| train on Full Training Set instead of Scene Flow only | MonSter++ · Zero-shot avg of 4 datasets | error rate: 3.70 -> 2.30 | Improved |
| train on Full Training Set instead of Scene Flow only | RT-MonSter++ · Zero-shot avg of 4 datasets | error rate: 7.03 -> 3.70 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- IGEV — IGEV++: Iterative Multi-Range Geometry Encoding Volumes for Stereo Matching (TPAMI 2025) · better on 7 of 7 shared results
- MVSFormer++ — MVSFormer++: Revealing the Devil in Transformer's Details for Multi-View Stereo (ICLR 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
- MonSter: Marry Monodepth to Stereo Unleashes Power (CVPR 2025 · similarity 0.96)
- Single View Stereo Matching (CVPR 2018 · similarity 0.91)
- MVS2D: Efficient Multi-View Stereo via Attention-Driven 2D Convolutions (CVPR 2022 · similarity 0.91)
- MVSAnywhere: Zero-Shot Multi-View Stereo (CVPR 2025 · similarity 0.90)
- Any-Stereo: Arbitrary Scale Disparity Estimation for Iterative Stereo Matching (AAAI 2024 · similarity 0.89)
- Multi-view Reconstruction via SfM-guided Monocular Depth Estimation (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.
