Research map / Detection and segmentation / CNN features
Template Matching via Densities on the Roto-Translation Group
Erik J. Bekkers, Marco Loog, Bart M. ter Haar Romeny, Remco Duits
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 SE(2) orientation-score templates instead of R2 image templates (average templates) (ablation) | template matching · ONH detection all images | success rate: 97.87% -> 99.08% | Improved |
| train templates by linear regression instead of averaging (SE(2)) (ablation) | SE(2) template · ONH detection all images | success rate: 99.08% -> 99.54% | Improved |
| add regularization to linear regression templates (SE(2)) (ablation) | linear regression SE(2) template · ONH detection all images | success rate: 94.01% -> 99.54% | Improved |
| add regularization to logistic regression templates (SE(2)) (ablation) | logistic regression SE(2) template · ONH detection all images | success rate: 83.14% -> 99.08% | Improved |
| add regularization to linear regression templates (R2) (ablation) | linear regression R2 template · ONH detection all images | success rate: 21.48% -> 97.07% | Improved |
| combine R2 and SE(2) templates (ablation) | template matching · ONH detection all images | success rate: 99.08% -> 99.83% | Improved |
| use logistic instead of linear regression for individual templates (ablation) | SE(2) templates · ONH detection all images | success rate: 99.54% -> 99.08% | Hurt |
| train templates instead of averaging in combinations (ablation) | template combination · ONH detection all images | success rate: 99.19% -> 99.83% | Improved |
| train SE(2) templates by linear regression instead of averaging (ablation) | SE(2) template · Fovea detection | success rate: 92.1% -> 94.6% | Improved |
| combine R2 and SE(2) templates (ablation) | template matching · Fovea detection | success rate: 94.6% -> 99.3% | Improved |
| use logistic instead of linear regression (ablation) | SE(2) template · Pupil detection | success rate: 87.2% -> 94.0% | Improved |
| train templates instead of averaging (ablation) | template matching · Pupil detection | success rate: 68.2% -> 95.6% | Improved |
Closest papers by meaning
- Tensorial template matching for fast cross-correlation with rotations and its application for tomography (ECCV 2024 · similarity 0.88)
- Template Matching With Deformable Diversity Similarity (CVPR 2017 · similarity 0.84)
- ReDiffDet: Rotation-equivariant Diffusion Model for Oriented Object Detection (CVPR 2025 · similarity 0.84)
- Arbitrary-Oriented Object Detection with Circular Smooth Label (ECCV 2020 · similarity 0.84)
- Selective Disk Bispectrum: A Complete and Rotation Invariant Image Descriptor (NeurIPS 2026 · similarity 0.83)
- Orient Anything V2: Unifying Orientation and Rotation Understanding (NeurIPS 2025 · similarity 0.83)
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.
