Computer vision engineer working on SLAM, visual-inertial odometry, 3D reconstruction and sensor fusion: from camera models and calibration to filters, bundle adjustment and dense mapping.
Background: M.Sc. EE (Ben-Gurion University). Worked on SLAM evaluation and ORB-SLAM at Magic Leap, edge stereo depth at Intel RealSense, and now multi-sensor mapping (stereo, LiDAR, 360°, RTK GNSS) on embedded GPUs.
| Repo | What it is |
|---|---|
| msckf_x | Monocular MSCKF VIO in C++, written from the papers: FEJ, fisheye/omni camera models, static + dynamic init, simulation gates, EuRoC/TUM-VI/KITTI benchmarks |
| indoor_3d | Offline indoor reconstruction from ARKit RGB-D: TSDF fusion, COLMAP pose refinement, depth upsampling |
| Repo | My changes |
|---|---|
| kalibr | Fixed-ξ omni (MEI) calibration with k2/k3 distortion, corner reuse from Basalt, target-pose seeding |
| basalt | Python bindings for project / unproject / RANSAC pose, spherical camera model for IMU↔360° calibration |
| basalt-headers | Equirectangular camera model for 360×180 panoramic input |
| opengv | SQPnP with angular error for omnidirectional cameras, Python wheels and CI |
| natural-neighbor-interpolation | CUDA support and a faster CPU path |
| pymcc / pymccrgb | Memory-leak and overflow fixes, faster surface interpolation for point-cloud ground classification |
Interests: 3D geometry · navigation · deep learning for geometry