Research map / 3D vision / Depth and stereo
Efficient Edge-Preserving Multi-View Stereo Network for Depth Estimation
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 CPC loss to the depth-domain loss (ablation) | EPNet without HEPR · DTU evaluation set | Overall distance (mm): 0.330 -> 0.327 | Helped |
| Add CPC loss to the depth-domain loss (ablation) | EPNet without HEPR · ETH3D training set | F1 (%): 63.84 -> 64.33 | Helped |
| Add the HEPR module for edge-preserving depth refinement and upsampling (ablation) | EPNet with CPC loss · DTU evaluation set | Overall distance (mm): 0.327 -> 0.311 | Helped |
| Add the HEPR module for edge-preserving depth refinement and upsampling (ablation) | EPNet with CPC loss · ETH3D training set | F1 (%): 64.33 -> 68.08 | Helped |
| Use five stages with hypotheses 32,16,8,8,8 instead of three stages with 32,16,8 (ablation) | EPNet MSDE · DTU evaluation set | Overall distance (mm): 0.339 -> 0.330 | Helped |
| Use five stages with hypotheses 32,16,8,8,8 instead of three stages with 32,16,8 (ablation) | EPNet MSDE · ETH3D training set | F1 (%): 59.22 -> 63.84 | Helped |
| Estimate the final depth at full image resolution rather than half resolution (ablation) | EPNet MSDE · DTU evaluation set and ETH3D training set | DTU Overall distance (mm) and ETH3D F1 (%): 0.339 -> 0.325; 59.22 -> 59.77 | Mixed |
| Train with CPC loss and HEPR in the full method (ablation) | EPNet · DTU evaluation set | Overall distance (mm): 0.330 -> 0.311 | Helped |
| Train with CPC loss and HEPR in the full method (ablation) | EPNet · ETH3D training set | F1 (%): 63.84 -> 68.08 | Helped |
Closest papers by meaning
- Rethinking Depth Estimation for Multi-View Stereo: A Unified Representation (CVPR 2022 · similarity 0.91)
- Cost Volume Pyramid Based Depth Inference for Multi-View Stereo (CVPR 2020 · similarity 0.90)
- Deep Stereo Using Adaptive Thin Volume Representation With Uncertainty Awareness (CVPR 2020 · similarity 0.90)
- GeoMVSNet: Learning Multi-View Stereo With Geometry Perception (CVPR 2023 · similarity 0.90)
- Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume (AAAI 2020 · similarity 0.90)
- A Global Depth-Range-Free Multi-View Stereo Transformer Network with Pose Embedding (NeurIPS 2024 · 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.