Research map / Label-efficient and robust learning / Recognition and long-tail
Convolutional Hough Matching Networks for Robust and Efficient Visual Correspondence
Juhong Min, Seungwook Kim, Minsu Cho
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 CHM layers (ablation) | CHMNet · SPair-71k | PCK@0.1: 21.6 -> 46.3 | Improved |
| use learnable local CHM instead of global RHM (ablation) | CHMNet · SPair-71k | PCK@0.1: 38.2 -> 46.3 | Improved |
| add last CHM layer (ablation) | CHMNet · SPair-71k | PCK@0.1: 43.1 -> 46.3 | Improved |
| add kernel G (ablation) | CHMNet · SPair-71k | PCK@0.1: 45.5 -> 46.3 | Improved |
| add soft sampler (ablation) | CHMNet · SPair-71k | PCK@0.1: 43.8 -> 46.3 | Improved |
| use semi-isotropic kernel instead of full kernel (4D-4D) (ablation) | CHM kernel · SPair-71k | PCK@0.1: 43.9 -> 44.5 | Improved |
| extend voting space from 4D to 6D (ablation) | psi kernel · SPair-71k | PCK@0.1: 44.5 -> 46.4 | Improved |
| use semi-isotropic kernel instead of full isotropic (4D-4D) (ablation) | CHM kernel · SPair-71k | PCK@0.1: 39.7 -> 44.5 | Improved |
| use full kernel instead of semi-isotropic (6D-4D) (ablation) | CHM kernel · SPair-71k | PCK@0.1: 46.4 -> 44.8 | Hurt |
| use sparse correlation with K=10 nearest neighbors (ablation) | CHMNet · SPair-71k | PCK@0.1: 46.4 -> 45.2 | No clear effect |
| increase scale-space resolution S from 3 to 7 (ablation) | CP-CHM CHMNet · SPair-71k | PCK@0.1 (F): 51.3 -> 51.6 | No clear effect |
| add center-pivot CHM kernels | CHMNet · per-pair inference time | time (ms): 54 -> 32 | Improved |
Closest papers by meaning
- Convolutional Hough Matching Networks (CVPR 2021 · similarity 0.99)
- Hyperpixel Flow: Semantic Correspondence With Multi-Layer Neural Features (ICCV 2019 · similarity 0.90)
- Convolutional Neural Network Architecture for Geometric Matching (CVPR 2017 · similarity 0.90)
- Universal Correspondence Network (NeurIPS 2016 · similarity 0.89)
- Deep Semantic Feature Matching (CVPR 2017 · similarity 0.88)
- Neural Matching Fields: Implicit Representation of Matching Fields for Visual Correspondence (NeurIPS 2022 · similarity 0.88)
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.
