Research map / 3D vision / Depth and stereo
Generalized Binary Search Network for Highly-Efficient Multi-View Stereo
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 dense depth regression with binary-search classification (ablation) | MVS depth estimation · DTU | Overall: 0.355 -> 0.360 | Helped |
| Add error tolerance bins to binary search (ablation) | GBi-Net · DTU | Overall: 0.360 -> 0.298 | Helped |
| Increase from 1 to 2 error tolerance bins on both sides (ablation) | GBi-Net · DTU | Overall: 0.298 -> 0.300 | Hurt |
| Increase from 1 to 2 error tolerance bins on both sides (ablation) | GBi-Net · DTU | <0.125 depth accuracy: 11.75 -> 13.58 | Helped |
| Increase from 1 to 2 error tolerance bins on both sides (ablation) | GBi-Net · DTU | Memory (MB): 2108 -> 2438 | Hurt |
| Use efficient gradient updating rather than accumulating gradients across stages (ablation) | GBi-Net · DTU | Training memory (MB): 12137 -> 5208 | Helped |
| Use efficient gradient updating rather than accumulating gradients across stages (ablation) | GBi-Net · DTU | Overall: 0.299 -> 0.298 | No clear effect |
| Remove random cropping data augmentation (ablation) | GBi-Net · DTU | Overall: 0.289 -> 0.298 | Hurt |
| Increase the number of search stages from 6 toward 8 (ablation) | GBi-Net · DTU | Reconstruction quality | Helped |
| Use GBi-Net instead of Bi-Net, adding error tolerance bins and the generalized search mechanisms (ablation) | MVS depth estimation · Tanks and Temples Advanced | Mean: 32.03 -> 37.32 | Helped |
Closest papers by meaning
- MVS2D: Efficient Multi-View Stereo via Attention-Driven 2D Convolutions (CVPR 2022 · similarity 0.93)
- Fast-MVSNet: Sparse-to-Dense Multi-View Stereo With Learned Propagation and Gauss-Newton Refinement (CVPR 2020 · similarity 0.93)
- Recurrent MVSNet for High-Resolution Multi-View Stereo Depth Inference (CVPR 2019 · similarity 0.91)
- Rethinking Depth Estimation for Multi-View Stereo: A Unified Representation (CVPR 2022 · similarity 0.90)
- MVSFormer++: Revealing the Devil in Transformer's Details for Multi-View Stereo (ICLR 2024 · similarity 0.90)
- MVSAnywhere: Zero-Shot Multi-View Stereo (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.