What problem does it solve? Building reliable robot perception requires solving recurring hard problems: calibrating cameras and LiDARs, streaming sensor data without frame drops or latency, synchronizing multiple sensors, and turning 2D detections into usable 3D positions. This Skill provides production-tested patterns and reference implementations for each of these tasks. ## Core Features & Use Cases - Camera Calibration: Intrinsic, extrinsic (stereo and camera-to-LiDAR), and hand-eye calibration workflows using OpenCV, with quality checklists and reprojection error validation. - Sensor Streaming: Thread-safe capture architecture with bounded buffers, capture-time timestamping, and diagnostics, plus multi-sensor timestamp synchronization and hardware trigger setup (RealSense inter-cam sync, PTP for GigE). - Perception Pipelines: Image undistortion with precomputed maps, object detection with depth-based 3D back-projection and workspace filtering, and fiducial marker (AprilTag/ArUco) pose estimation. - Use Case: When integrating a RealSense or ZED camera into a ROS2 robot, use this Skill to calibrate intrinsics, set up a drop-free streaming thread, and convert detections into 3D points for manipulation or navigation. ## Quick Start Use the robot-perception skill to calibrate my camera intrinsics from checkerboard images and set up a synchronized RGB-depth streaming pipeline.
