Why I built it
Spatial capture can appear functional while hiding cadence, energy, thermal, and data-integrity costs on the target device.
What exists
The research harness exercises multiple sensing modes; performance and thermal causality remain open pending controlled device measurements.
How it works
AR callbacks drive camera, depth, mesh, relocalization, and protected capture paths with explicit instrumentation needs.
- Scene depth and mesh research
- Relocalization experiments
- Protected capture paths
- Performance-gate design
My role
Audited the native pipeline and defined controlled measurement and thermal-policy gates.
What I checked
- source-reviewNative capture and cadence path audited
What stays bounded
- No thermal cause is claimed from uncontrolled runs
- Illustrative diagrams are not measured geometry
What is missing
- Controlled cooled, unplugged target-device baselines are incomplete.
Next
Complete controlled target-device performance baselines.
Later ideas, not current features
- Run controlled profiling
- Add explicit resting and thermal states