Mission Turn synchronised recordings from a six-camera headset into accurate 3D data: camera poses, hand and body poses, and scene geometry.
Responsibilities Calibrate the multi-camera rig (intrinsics/extrinsics) and validate calibration over time. Estimate head/camera trajectory (visual or visual-inertial odometry, SLAM). Run and adapt 3D hand and body pose estimation from multi-view footage. Measure accuracy and report failure cases.
Must have Solid grounding in multi-view geometry and camera calibration. Python, PyTorch, OpenCV. Hands-on project or publication in 3D vision, pose estimation or SLAM.
Nice to have Experience with hand pose models (e.g. MANO-based methods). COLMAP, ORB-SLAM or similar tools. Egocentric vision research
