straintrace --help and python -m ezdic_cli --help expose the same interface.
straintrace campaign --help adds a separate version-1 campaign interface for
discovery, explicit ROI approval, resumable analysis, mechanics alignment and
versioned delivery. The existing single-sequence configuration remains unchanged.
See campaign workflow for input identities, missing values,
recorded versus derived clocks and imported-result preservation.
validate-config --config run.json: validate and print normalized version-1 JSON.run --config run.json --progress-json: execute a permitted image sequence.verify-manifest --manifest PATH: verify the recorded output bundle (see--helpfor syntax).benchmark --output PATH: run the locked, generated verification cases.
The schema is in schemas/run_config_v1.json in source and straintrace_data in the
wheel. The README contains complete extensometer and full-field configuration examples.
Unknown fields and non-finite configuration numbers are rejected.
Full-field configurations additionally accept mask (none/auto/file plus
reference-image exclusion rectangles) and display (contour/points, explicit
color range, percent units, colormap and reference/deformed background).
Solver options include initialization (auto/local), strain_degree (1/2),
robust_strain, and outlier_threshold_px (zero disables spatial rejection).
The legacy strain_window_px counts POIs, not image pixels. See
the 2D workflow for scientific interpretation.
Python interfaces: ezdic_cli.normalize_config(mapping, base_dir=path) returns the
canonical configuration; ezdic_cli.main(argv) returns a process exit status;
ezdic_benchmark.run_benchmark(output_dir=path) returns a machine-readable report.
ezdic_core contains the GUI-independent numerical implementation. Private helpers
are not a stability promise; prefer the versioned JSON interface for integration.
Coordinates and displacement are in pixels. Strain is dimensionless; tensor shear must not be interpreted as engineering shear. Retain both displacement-valid and strain-valid indicators. A verified checksum establishes integrity, not measurement accuracy. A successful synthetic benchmark does not calibrate experimental uncertainty.