Research map / 3D vision / Depth and stereo
Two-Way Street: Efficient VSLAM using Collaborative In-Sensor and Off-Sensor processing
Hongyi Zhang, Laurie Bose, Piotr Dudek, Walterio Mayol-Cuevas
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 |
|---|---|---|---|
| move feature detection and tracking into the sensor (PPA) instead of a RealSense camera | ORB SLAM 3 · Back-end tracking thread | frame time (us): ≈14, 834 -> ≈1, 554 | Improved |
| move feature detection and tracking into the sensor (PPA) | ORB SLAM 3 · Back-end mapping thread | frame time (us): ≈87, 927 -> ≈6, 856 | Improved |
| process per frame in sensor at 2000 FPS instead of RealSense at 60 FPS | ORB SLAM 3 · Energy per frame | energy (% battery): (4.53 ± 0.14) × 10−4 -> (1.9 ± 0.1) × 10−5 | Improved |
| process in sensor instead of RealSense | ORB SLAM 3 · Energy per second | energy (% battery): (2.35 ± 0.08) × 10−2 -> (2.34 ± 0.07) × 10−2 | No clear effect |
| send sparse features instead of whole images | ORB SLAM 3 · Data per frame | bytes: > 500,000 -> ≈200 | Improved |
| use in-sensor tracking with pose feedback instead of camera ORB SLAM 3 | ORB SLAM 3 · Rapid-motion sequences | times lost tracking: 80.2 -> 2.2 | Improved |
| use in-sensor tracking instead of camera | ORB SLAM 3 · Rapid-motion sequences | frames with no tracking (%): 9.06 -> 1.44 | Improved |
| use in-sensor tracking instead of camera | ORB SLAM 3 · Rapid motion translation | APE rmse (m): 0.113 -> 0.046 | Improved |
| use in-sensor tracking instead of camera | ORB SLAM 3 · Rapid motion rotation | RPE rmse (deg): 4.902 -> 1.089 | Improved |
Closest papers by meaning
- Event-based Simultaneous Localization and Mapping: A Comprehensive Survey (IJCV 2026 · similarity 0.88)
- PLGSLAM: Progressive Neural Scene Represenation with Local to Global Bundle Adjustment (CVPR 2024 · similarity 0.84)
- Descriptor-In-Pixel : Point-Feature Tracking For Pixel Processor Arrays (CVPR 2025 · similarity 0.84)
- Continuous Pose for Monocular Cameras in Neural Implicit Representation (CVPR 2024 · similarity 0.83)
- 360Loc: A Dataset and Benchmark for Omnidirectional Visual Localization with Cross-device Queries (CVPR 2024 · similarity 0.83)
- Swift-Mapping: Online Neural Implicit Dense Mapping in Urban Scenes (AAAI 2024 · 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.
