Research map / 3D vision / Depth and stereo
SurfaceNet+: An End-to-end 3D Neural Network for Very Sparse Multi-View Stereopsis
Mengqi Ji, Jinzhi Zhang, Qionghai Dai, Lu Fang
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 multi-scale coarse-to-fine inference (ablation) | SurfaceNet · DTU model 23 (sparsity 5) | precision | Improved |
| add trainable occlusion-aware view selection (ablation) | SurfaceNet multi-scale · DTU model 23 (sparsity 5) | recall and precision | Improved |
| add trainable relative weight with occlusion awareness (ablation) | SurfaceNet+ · DTU model 23 (sparsity 5) | recall | Improved |
| trainable view selection instead of heuristic view selection (ablation) | SurfaceNet+ · DTU model 23 (sparsity 5) | recall | Improved |
| add multi-scale feature aggregation (ablation) | SurfaceNet+ · DTU model 23 (sparsity 5) | recall vs precision | Mixed |
| replace ICCV SurfaceNet backbone with the new backbone (ablation) | SurfaceNet · DTU model 23 (sparsity 5) | recall and precision | No clear effect |
| retrain R-MVSNet on randomly sampled sparse view pairs | R-MVSNet · DTU (Batchsize 1) | f-score <1mm | No clear effect |
| add coarse-to-fine volume selection (ablation) | SurfaceNet · DTU | mean speed up: 11X | Improved |
| add coarse-to-fine volume selection (ablation) | SurfaceNet · Tanks and Temples | mean speed up: 15X | Improved |
| use noisy SfM camera poses instead of GT poses | SurfaceNet+ and baselines · DTU | f-score | Hurt |
| larger sparsity with occlusion parameter alpha (more recall gain) | SurfaceNet+ · DTU | recall gain | Improved |
Later papers that compared with this method
1 later accepted paper reports SurfaceNet as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Dense Hybrid Recurrent Multi-view Stereo Net with Dynamic Consistency Checking (ECCV 2020 · 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
- Just a Few Points Are All You Need for Multi-View Stereo: A Novel Semi-Supervised Learning Method for Multi-View Stereo (ICCV 2021 · similarity 0.90)
- Stereo Radiance Fields (SRF): Learning View Synthesis for Sparse Views of Novel Scenes (CVPR 2021 · similarity 0.90)
- Fast-MVSNet: Sparse-to-Dense Multi-View Stereo With Learned Propagation and Gauss-Newton Refinement (CVPR 2020 · similarity 0.90)
- Learning a Multi-View Stereo Machine (NeurIPS 2017 · similarity 0.90)
- MVS2D: Efficient Multi-View Stereo via Attention-Driven 2D Convolutions (CVPR 2022 · similarity 0.90)
- DeepMVS: Learning Multi-View Stereopsis (CVPR 2018 · similarity 0.90)
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.
