Submitted anonymously.

Submission data

Full name	RobustSparseGeometryMapping
Description	A sparse geometric mapping framework that estimates camera motion and reconstructs 3D landmarks from stereo image pairs. The method employs robust outlier rejection, adaptive landmark selection, and graph-based optimization to maintain accurate localization in large-scale environments. IMU integration improves motion estimation during rapid movements and low-texture scenes.
Parameters	Maximum landmarks: 5000 Feature detection threshold: 0.02 RANSAC iterations: 1000 Keyframe insertion distance: 0.5 m Graph optimization frequency: every 20 frames IMU fusion weight: 0.40
Programming language(s)	C++, Eigen, OpenCV
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