Submitted anonymously.

Submission data

Full name	AdaptiveHybridVisualOdometry
Description	A hybrid visual odometry system that combines direct image alignment with sparse feature tracking. The method dynamically switches between tracking modes based on scene texture and motion characteristics. Stereo depth and IMU measurements are fused through an adaptive Kalman filter, while loop closure detection reduces long-term drift and improves global consistency.
Parameters	Maximum tracked features: 3000 Direct alignment pyramid levels: 4 Feature re-detection interval: 10 frames IMU update rate: 200 Hz Loop closure similarity threshold: 0.85 Pose optimization iterations: 50
Programming language(s)	C++, Python, CUDA
Hardware	A100
Submission creation date	23 Jun, 2026
Last edited	23 Jun, 2026

High-res multi-view results

Low-res many-view results

Low-res two-view results

SLAM results