Research map / 3D vision / Depth and stereo
RC-MVSNet: Unsupervised Multi-View Stereo with Neural Rendering
Di Chang, Aljaž Božič, Tong Zhang, Qingsong Yan, Yingcong Chen, Sabine Süsstrunk, Matthias Nießner
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 data augmentation loss L_DA (ablation) | RC-MVSNet baseline · DTU | Overall: 0.395 -> 0.383 | Improved |
| add reference view synthesis loss L_RC (ablation) | RC-MVSNet · DTU | Overall: 0.383 -> 0.357 | Improved |
| add Gaussian-Uniform mixture sampling (ablation) | RC-MVSNet with L_RC · DTU | Overall: 0.357 -> 0.352 | Improved |
| add depth rendering consistency loss L_DC (ablation) | RC-MVSNet with L_RC · DTU | Overall: 0.357 -> 0.348 | Improved |
| add L_DC on top of Gaussian-Uniform sampling (ablation) | RC-MVSNet · DTU | Overall: 0.352 -> 0.345 | Improved |
| increase sampled rays from 256 to 1024, 4096, 8192 (ablation) | RC-MVSNet · DTU | Overall: 0.350 -> 0.345 (1024), 0.350 (4096), 0.348 (8192) | No clear effect |
| RC-MVSNet instead of JDACS-MS | unsupervised E2E MVS · DTU | Overall: 0.358 -> 0.345 | Improved |
| RC-MVSNet instead of JDACS-MS | unsupervised E2E MVS · DTU | Completeness: 0.318 -> 0.295 | Improved |
| RC-MVSNet instead of M3VSNet | unsupervised E2E MVS · Tanks and Temples intermediate | mean F-score: +17.37 | Improved |
| RC-MVSNet instead of JDACS-MS | unsupervised E2E MVS · Tanks and Temples intermediate | mean F-score: 45.48 -> 55.04 | Improved |
Later papers that compared with this method
3 later accepted papers report RC-MVSNet as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- "Smoothness, Synthesis, and Sampling: Re-thinking Unsupervised Multi-View Stereo with DIV Loss" (ECCV 2024 · better on 3 of 3 shared results)
- Surface-Centric Modeling for High-Fidelity Generalizable Neural Surface Reconstruction (ECCV 2024 · better on 1 of 1 shared result)
- CL-MVSNet: Unsupervised Multi-View Stereo with Dual-Level Contrastive Learning (ICCV 2023 · better on 4 of 4 shared results)
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.
- PatchmatchNet — PatchmatchNet: Learned Multi-View Patchmatch Stereo (CVPR 2021) · better on 2 of 3 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
- CL-MVSNet: Unsupervised Multi-View Stereo with Dual-Level Contrastive Learning (ICCV 2023 · similarity 0.92)
- "Smoothness, Synthesis, and Sampling: Re-thinking Unsupervised Multi-View Stereo with DIV Loss" (ECCV 2024 · similarity 0.91)
- Multi-View Stereo Representation Revist: Region-Aware MVSNet (CVPR 2023 · similarity 0.91)
- S-VolSDF: Sparse Multi-View Stereo Regularization of Neural Implicit Surfaces (ICCV 2023 · similarity 0.89)
- MVSFormer++: Revealing the Devil in Transformer's Details for Multi-View Stereo (ICLR 2024 · similarity 0.89)
- MVCPS-NeuS: Multi-view Constrained Photometric Stereo for Neural Surface Reconstruction (CVPR 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.
