Research map / 3D vision / Depth and stereo
WT-MVSNet: Window-based Transformers for Multi-view Stereo
Read the paperCheck 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 |
|---|---|---|---|
| Add CE Loss to the baseline (ablation) | CasMVSNet-based baseline · DTU evaluation | Overall (mm): 0.331 -> 0.325 | Helped |
| Add Geo Loss (ablation) | Baseline with CE Loss · DTU evaluation | Overall (mm): 0.325 -> 0.319 | Helped |
| Add WET (ablation) | Baseline with CE Loss and Geo Loss · DTU evaluation | Overall (mm): 0.319 -> 0.308 | Helped |
| Add CT (ablation) | Baseline with CE Loss, Geo Loss, and WET · DTU evaluation | Overall (mm): 0.308 -> 0.295 | Helped |
| Replace no cost regularization with CT (ablation) | WT-MVSNet · DTU evaluation | Overall (mm): 0.422 -> 0.295 | Helped |
| Replace 3D CNN regularization with CT (ablation) | WT-MVSNet · DTU evaluation | Overall (mm): 0.308 -> 0.295 | Helped |
| Increase WET attention blocks from 1 to 2 (ablation) | WT-MVSNet · DTU evaluation | Overall (mm): 0.295 -> 0.295 | No clear effect |
| Increase WET attention blocks from 1 to 3 (ablation) | WT-MVSNet · DTU evaluation | Overall (mm): 0.295 -> 0.298 | Hurt |
| Change WET window from 16×16 to 24×24 (ablation) | WT-MVSNet · DTU evaluation | Overall (mm): 0.295 -> 0.304 | Hurt |
| Increase CT attention blocks from 1 to 3 (ablation) | WT-MVSNet · DTU evaluation | Overall (mm): 0.303 -> 0.295 | Helped |
| Set Geo Loss weight λ₂ to 1 instead of 0 (ablation) | WT-MVSNet · DTU evaluation | Overall (mm): 0.302 -> 0.295 | Helped |
| Increase Geo Loss weight λ₂ from 1 to 2 (ablation) | WT-MVSNet · DTU evaluation | Overall (mm): 0.295 -> 0.297 | Hurt |
Closest papers by meaning
- RRT-MVS: Recurrent Regularization Transformer for Multi-View Stereo (AAAI 2025 · similarity 0.88)
- A Global Depth-Range-Free Multi-View Stereo Transformer Network with Pose Embedding (NeurIPS 2024 · similarity 0.88)
- MVSFormer++: Revealing the Devil in Transformer's Details for Multi-View Stereo (ICLR 2024 · similarity 0.87)
- RayMVSNet: Learning Ray-Based 1D Implicit Fields for Accurate Multi-View Stereo (CVPR 2022 · similarity 0.87)
- Understanding Multi-View Transformers (NeurIPS 2026 · similarity 0.85)
- TurboVGGT: Fast Visual Geometry Reconstruction with Adaptive Alternating Attention (NeurIPS 2026 · similarity 0.85)
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.