Technical Field Notes
Measured Capture Evidence
Recomputed sensor results, timing models, calibration findings and stated limitations from Origin Data Lab captures
Each note reports what was measured from recorded timestamps and raw sensor files, how it was measured, and what the evidence does not claim. Notes are written for robotics, VIO and Physical AI engineers evaluating real-world capture pipelines.
All Notes
Latest Field Notes
Checking a Stereo Calibration on Ordinary Footage
A target-free check for fisheye stereo calibrations: 0.41 px on a mode-matched calibration vs 1.83 px on a mismatched one, 0.1° sensitivity, and two metric traps that make good calibrations look bad.
FIELD NOTE #03Stereo + IMU Egocentric Capture: Synchronization, Timing and Calibration
3,600 synchronized stereo pairs with per-frame exposure timing, 501.7 Hz IMU, a target-free camera–IMU offset estimate, and why stereo calibration must match the sensor mode.
FIELD NOTE #02Egocentric RGB + IMU Task Sample
First-party egocentric food-preparation capture with measured sensor rates, software-derived video–IMU alignment, keystep task labels, integrity checks and a public evaluation package.
FIELD NOTE #01Measured Video + IMU Capture Characteristics
12 consecutive segments of a continuous real-world produce-sorting recording: effective video FPS, ~410 Hz native 6-DoF IMU, timestamp integrity and temporal coverage, with timing gaps reported.
Method
How We Report
Technical Validation
Need Evidence for Your Own Capture Requirement?
We define capture architecture, sensor timing, metadata, validation criteria and structured delivery around your robotics or Physical AI task.