Research map / 3D vision / Depth and stereo
PDC-Net+: Enhanced Probabilistic Dense Correspondence Network
Prune Truong, Martin Danelljan, Radu Timofte, Luc Van Gool
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 |
|---|---|---|---|
| use PDC-Net+ training strategy (multiple objects, injective mask) instead of PDC-Net | PDC-Net (D) · KITTI-2015 | Fl (%): 15.13 -> 12.62 | Improved |
| same change on KITTI-2012 | PDC-Net (D) · KITTI-2012 | AEPE: 2.08 -> 1.76 | Improved |
| PDC-Net+ vs PDC-Net on MegaDepth | PDC-Net (H) · MegaDepth | PCK-1 (%): 70.75 -> 73.92 | Improved |
| add probabilistic formulation (PDC-Net+ vs non-probabilistic baseline) | GLU-Net-GOCor* · MegaDepth | PCK-1 (%): 57.77 -> 72.41 | Improved |
| use multi-scale (MS) inference instead of single pass (D) | PDC-Net+ · MegaDepth | PCK-1 (%): 72.41 -> 74.51 | Improved |
| use multi-scale inference instead of single pass | PDC-Net+ · RobotCar | PCK-1 (%): 2.57 -> 2.63 | No clear effect |
| add multi-scale strategy to PDC-Net+ (H) | PDC-Net+ · HPatches | AEPE: 4.29 -> 3.59 | Improved |
| PDC-Net+ vs PDC-Net | PDC-Net (H) · HPatches | AEPE: 4.32 -> 4.29 | No clear effect |
| use confidence P_R instead of mixture variance (ablation) | PDC-Net+ (D) · MegaDepth | AUSE PCK-5: 0.0171 -> 0.0112 | Improved |
| use confidence P_R instead of forward-backward consistency (ablation) | PDC-Net+ (D) · MegaDepth | AUSE PCK-5: 0.0266 -> 0.0112 | Improved |
| use confidence P_R instead of cyclic error (ablation) | PDC-Net+ (D) · MegaDepth | AUSE AEPE: 0.2953 -> 0.2482 | Improved |
| fix sigma1 squared to 1.0 instead of letting it vary in [0,1] (ablation) | PDC-Net+ · text | performance | Improved |
Closest papers by meaning
- Learning Accurate Dense Correspondences and When To Trust Them (CVPR 2021 · similarity 0.93)
- Dense Flow from Adaptive Correspondence (NeurIPS 2026 · similarity 0.87)
- Deep Matching Prior: Test-Time Optimization for Dense Correspondence (ICCV 2021 · similarity 0.87)
- DKM: Dense Kernelized Feature Matching for Geometry Estimation (CVPR 2023 · similarity 0.87)
- UFM: A Simple Path towards Unified Dense Correspondence with Flow (NeurIPS 2025 · similarity 0.85)
- Diffusion Model for Dense Matching (ICLR 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.
