Research Experiments

Phase 10 camera streams, proxy/original choice, synchronization transform, orientation, calibration, adapter/parser versions, and source hashes become pinnable experimental inputs. Changing them makes dependent interpretation stale rather than rewriting an experiment.

Phase 9 provides versioned, resumable experiments over pinned datasets and annotation revisions. Experiments move through draft, validation, queued, running, paused, completed, warning, failure, cancellation, interruption, stale, and invalid states.

Every result records data origin, pipeline and plugin versions, device details, configuration, integrity hashes, warnings, measured metrics, and abstentions. Deterministic simulator results demonstrate architecture and reproducibility only; they do not establish real-world performance.

Use the Experiment Browser to create, duplicate, compare, pause, resume, cancel, export, or inspect an experiment. Phase 9 does not train or deploy learned models.