Research map / 3D vision / Depth and stereo
EPP-MVSNet: Epipolar-Assembling Based Depth Prediction for Multi-View Stereo
Xinjun Ma, Yue Gong, Qirui Wang, Jingwei Huang, Lei Chen, Fan Yu
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 epipolar-assembling module at the same depth number of 32 (ablation) | EPP-MVSNet 1st stage · DTU evaluation | overall error (mm): 0.7049 -> 0.5920 | Improved |
| add epipolar-assembling instead of raising depth number to 96 (ablation) | EPP-MVSNet 1st stage · DTU evaluation | overall error (mm): 0.6282 -> 0.5920 | Improved |
| add epipolar-assembling instead of raising depth number to 160 (ablation) | EPP-MVSNet 1st stage · DTU evaluation | overall error (mm): 0.6298 -> 0.5920 | Improved |
| use max pooling instead of average pooling in assembling kernel (ablation) | EAM kernel · DTU evaluation | overall error (mm): 0.5994 -> 0.5920 | Improved |
| use 3x1x1 convolution kernel instead of 1x1x1 (ablation) | EAM kernel · DTU evaluation | overall error (mm): 0.7163 -> 0.5920 | Improved |
| use 5x1x1 kernel instead of 3x1x1 (ablation) | EAM kernel · DTU evaluation | overall error (mm): 0.5920 -> 0.5921 | No clear effect |
| add entropy-based refining strategy (ablation) | EPP-MVSNet 3rd stage · DTU evaluation | overall error (mm): 0.3595 -> 0.3553 | Improved |
| replace 3D CNN with Pseudo-3D CNN in regularization (ablation) | EPP-MVSNet · DTU evaluation | overall error (mm): 0.3575 -> 0.3553 | No clear effect |
| replace 3D CNN with Pseudo-3D CNN in regularization (ablation) | EPP-MVSNet · DTU evaluation | time (ms): 624.9 -> 340.6 | Improved |
| replace 3D CNN with Pseudo-3D CNN in regularization (ablation) | EPP-MVSNet · DTU evaluation | memory (GB): 5.2 -> 3.1 | Improved |
| replace 3D block with Pseudo-3D block on 32x296x400 volume (ablation) | regularization block · DTU evaluation | run-time (ms): 23.6 -> 16.4 | Improved |
| replace 3D block with Pseudo-3D block on 64x296x400 volume (ablation) | regularization block · DTU evaluation | run-time (ms): 46.6 -> 31.7 | Improved |
Later papers that compared with this method
5 later accepted papers report EPP-MVSNet as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- Adaptive Patch Deformation for Textureless-Resilient Multi-View Stereo (CVPR 2023 · better on 1 of 1 shared result)
- TransMVSNet: Global Context-Aware Multi-View Stereo Network With Transformers (CVPR 2022 · better on 4 of 4 shared results)
- Generalized Binary Search Network for Highly-Efficient Multi-View Stereo (CVPR 2022 · better on 1 of 2 shared results)
- Rethinking Depth Estimation for Multi-View Stereo: A Unified Representation (CVPR 2022 · better on 1 of 1 shared result)
- MVSTER: Epipolar Transformer for Efficient Multi-View Stereo (ECCV 2022 · better on 1 of 1 shared result)
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 3 of 5 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
- MVSTER: Epipolar Transformer for Efficient Multi-View Stereo (ECCV 2022 · similarity 0.89)
- RayMVSNet: Learning Ray-Based 1D Implicit Fields for Accurate Multi-View Stereo (CVPR 2022 · similarity 0.89)
- When Epipolar Constraint Meets Non-Local Operators in Multi-View Stereo (ICCV 2023 · similarity 0.88)
- P-MVSNet: Learning Patch-Wise Matching Confidence Aggregation for Multi-View Stereo (ICCV 2019 · similarity 0.88)
- Efficient Edge-Preserving Multi-View Stereo Network for Depth Estimation (AAAI 2023 · similarity 0.87)
- A Global Depth-Range-Free Multi-View Stereo Transformer Network with Pose Embedding (NeurIPS 2024 · 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.
