Document the extraction.
Keep the limitations visible.

Four anchors, two linear axes

Select two points with known values on the displacement axis, and two on the force axis. The conversion resolves each pixel into the two axis directions, then applies their value scales. Axis lines can be rotated or sheared; anchors with zero span or nearly parallel directions are rejected.

Use straight, linear axes. Logarithmic scales, broken axes, perspective distortion and photographs of skewed pages require preparation outside this tool. Zoom changes only the display. Original image pixels remain the calibration coordinates.

Displacement units

mm, cm, m or in are converted to mm. One inch equals 25.4 mm.

Force units

kN, N or lbf are converted to kN. One lbf equals 0.0044482216152605 kN.

Trace order

Points retain the order you enter. Cyclic loops and displacement reversals are not sorted into a monotonic curve.

Interpolation

Lines between points are piecewise linear. They do not restore unrecorded detail or original sampling frequency.

Comparison with a model

Import a model CSV, choose its displacement and force columns, and state its units, protocol and measurement definitions. Confirm that both records represent comparable experimental conditions before using error metrics.

Calculations use the existing curve comparison methods. Capacity is the maximum absolute recorded force. Secant stiffness uses the 10–40% ascending branch only when its definition is applicable. Monotonic RMSE is integrated over the shared displacement interval without extrapolation; displacement coverage is always reported with it. Cyclic traces are not collapsed into a single-valued function.

All comparison values use the digitized approximation as their reference. A close match does not establish predictive validity, code compliance or design suitability. Missing or incompatible measurement definitions withhold error metrics.

Traceability and reproducible packages

A saved project records the source citation and locator, specimen description, source-image fingerprint, axis anchors, units, ordered pixel points and normalized curve. The project ZIP includes the original imported image bytes. Reopening a project checks the image fingerprint and recalculates the curve from the trace.

Edits increase the current project revision. A saved package is a snapshot of that revision; it is not an immutable log of every earlier edit. Keep earlier ZIPs when you need a full sequence of extraction versions. CSV exports contain displacement in mm and force in kN.

Image fingerprints identify matching bytes; they do not prove authorship, permission, accuracy or an unchanged external publication. Processing runs locally. Work is kept in the current browser tab until you download a package; closing the tab can lose unsaved changes.

Source review and sharing

Record the exact page, figure and specimen, measurement definitions, geometry, loading, connections, limitations and reuse terms. Confirm permission before submitting a contribution. Public access to a paper alone does not establish permission to redistribute its figures or data.

An explicit submission sends the structured extraction record and image fingerprint through an approved academic account. The image is not uploaded; it remains in the local ZIP. The owner must inspect the cited source figure and review the record. Review notes are visible to the submitter.

Approval records source metadata and sharing review. It does not independently validate the experiment or the tracing. A separate publication action releases contribution metadata and credit only; source pixels and the private structured payload are not made public. Atlas specimens and Exchange cases require separate owner curation.

Extraction accuracy

Image resolution, line thickness, overlapping curves, calibration choices and point placement all affect digitization. The Builder does not assign a numerical uncertainty automatically. Check selected coordinates against the original figure, include turning points and document the extraction method before reusing the data.