Research map / Architectures and efficiency / CNN design
Iterative Continuous Convolution for 3D Template Matching and Global Localization
Vitor Guizilini, Fabio Ramos
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 |
|---|---|---|---|
| Replace GEN-ICP with CC-HM | 3D template matching · Corridor | computation time (s): 55.9 -> 2.9 | Improved |
| Replace GEN-ICP with CC-HM | 3D template matching · Corridor | linear error (cm): 11.9 -> 14.5 | Hurt |
| Replace GEN-ICP with CC-HM | 3D template matching · Outdoor | linear error (cm): 15.5 -> 17.1 | Hurt |
| Add iterative convolution (ICC-HM) on top of CC-HM (ablation) | CC-HM · Corridor | linear error (cm): 14.5 -> 9.8 | Improved |
| Add iterative convolution (ICC-HM) on top of CC-HM (ablation) | CC-HM · Outdoor | linear error (cm): 17.1 -> 10.9 | Improved |
| Add iterative convolution (ICC-HM) on top of CC-HM (ablation) | CC-HM · Corridor | computation time (s): 2.9 -> 4.1 | Hurt |
| Add iterative convolution (ICC-HM) on top of CC-HM (ablation) | CC-HM · Corridor | angular error (deg): 4.4 -> 3.1 | Improved |
| Use ICC-HM output to initialize GEN-ICP (ICC-ICP) (ablation) | ICC-HM · Corridor | linear error (cm): 9.8 -> 9.5 | Improved |
| Use ICC-HM output to initialize GEN-ICP (ICC-ICP) (ablation) | ICC-HM · Outdoor | computation time (s): 5.7 -> 9.3 | Hurt |
| Continuous convolution instead of Fourier convolution at 0.02 m | convolution · Convolution timing | inference time (us/pt): 5.1 -> 1.0 | Improved |
| Continuous convolution instead of sliding convolution at 0.02 m | convolution · Convolution timing | grid generation time (ms/m3): 167.9 -> 2.5 | Improved |
| Use continuous convolution at very small mask sizes | convolution · speed vs mask size | speed | Hurt |
Closest papers by meaning
- Efficient Global 2D-3D Matching for Camera Localization in a Large-Scale 3D Map (ICCV 2017 · similarity 0.86)
- DH3D: Deep Hierarchical 3D Descriptors for Robust Large-Scale 6DoF Relocalization (ECCV 2020 · similarity 0.86)
- Efficient & Effective Prioritized Matching for Large-Scale Image-Based Localization (TPAMI 2016 · similarity 0.85)
- Templates for 3D Object Pose Estimation Revisited: Generalization to New Objects and Robustness to Occlusions (CVPR 2022 · similarity 0.85)
- Is Geometry Enough for Matching in Visual Localization? (ECCV 2022 · similarity 0.85)
- LOCATE 3D: Real-World Object Localization via Self-Supervised Learning in 3D (ICML 2025 · similarity 0.84)
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.
