Research map / 3D vision / Human pose estimation
Cascaded Parallel Filtering for Memory-Efficient Image-Based Localization
Wentao Cheng, Weisi Lin, Kan Chen, Xinfeng Zhang
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 bilateral Hamming ratio test with two-step match selection (vs baseline voting) (ablation) | Our method · Dubrovnik | localized images: 778 -> 794 | Improved |
| add bilateral Hamming ratio test with two-step match selection (vs baseline voting) (ablation) | Our method · Dubrovnik | median error (m): 0.69 -> 0.64 | Improved |
| add quality-aware spatial reconfiguration (QSR) (ablation) | Our method · Dubrovnik | median error (m): 0.74 -> 0.64 | Improved |
| add quality-aware spatial reconfiguration (QSR) (ablation) | Our method · Dubrovnik | 75% error quartile (m): 2.53 -> 2.16 | Improved |
| add quality-aware spatial reconfiguration (QSR) (ablation) | Our method · Dubrovnik | localized images: 793 -> 794 | No clear effect |
| add principal focal length estimation (PFL) (ablation) | Our method · Dubrovnik | median error (m): 0.80 -> 0.64 | Improved |
| add principal focal length estimation (PFL) (ablation) | Our method · Dubrovnik | 75% error quartile (m): 2.70 -> 2.16 | Improved |
| store compact binary signatures instead of integer mean SIFT descriptors (vs Active Search) | Our method · Dubrovnik | memory (GB): 0.75 -> 0.14 | Improved |
| store compact binary signatures instead of integer mean SIFT descriptors (vs Active Search) | Our method · RobotCar Seasons | memory (GB): 2.72 -> 0.52 | Improved |
| store compact binary signatures instead of integer mean SIFT descriptors (vs Active Search) | Our method · Aachen Day-Night | memory (GB): 0.76 -> 0.14 | Improved |
| use cascaded filtering without gravity prior vs CSL with gravity prior | Our method · Aachen Day-Night (day, 0.25m/2deg) | localized %: 52.3 -> 76.7 | Improved |
| use no semantic network vs SMC with semantic segmentation | Our method · RobotCar Seasons (all night, 0.25m/2deg) | localized %: 7.6 -> 3.4 | Hurt |
Closest papers by meaning
- Efficient & Effective Prioritized Matching for Large-Scale Image-Based Localization (TPAMI 2016 · similarity 0.88)
- Improving Transformer-based Image Matching by Cascaded Capturing Spatially Informative Keypoints (ICCV 2023 · similarity 0.86)
- Combining Absolute and Semi-Generalized Relative Poses for Visual Localization (IJCV 2026 · similarity 0.86)
- Learning To Detect Scene Landmarks for Camera Localization (CVPR 2022 · similarity 0.85)
- RELOCATE: A Simple Training-Free Baseline for Visual Query Localization Using Region-Based Representations (CVPR 2025 · similarity 0.84)
- Reloc3r: Large-Scale Training of Relative Camera Pose Regression for Generalizable, Fast, and Accurate Visual Localization (CVPR 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.
