Scientific Graph Compare

Digitised points are estimates read off a picture, not measurements Everything here inherits the error already baked into the published figure, plus your calibration clicks on top. Values taken this way are an approximation of what a plot shows, never a substitute for the author's underlying data. Open for the full scope limits.

What this tool actually does

Maps pixel positions to data coordinates using the axis calibration you provide, on a linear or log10 scale, and exports the points you place. The pixel-to-data transform itself is tested against known references.

The transform being right is not the same as the numbers being right.

Where the error comes from

Your calibration clicks. Every point inherits the error in the two axis references you set. Placing them close together magnifies it across the whole plot; place them as far apart as the axes allow.
Log axes are not forgiving. On a log scale a fixed pixel error is a constant relative error, so a click near the low end of a decade carries far more absolute error than the same click near the top. Choosing linear when the axis is logarithmic, or the reverse, produces smooth, plausible, completely wrong values.
The marker is not the datum. A symbol has width. Its centroid, its edge and the value it represents are three different points, and thick lines, overlapping series and anti-aliasing all move your click.
The figure is already lossy. A published plot has been rounded to its axis resolution, rasterised, compressed and possibly rescaled before you ever saw it. Digitising cannot recover precision the figure does not contain.
Not captured at all: error bars, uncertainties, sample sizes, censored or excluded points, axis breaks, secondary axes, and any footnote that changes what the plot means.
Distortion. Photographed, scanned, skewed or perspective-shifted figures need rectifying first, and a plot that is not square-on will read consistently wrong.

Report digitised values as approximate, say they were digitised, and cite the original figure. Presenting them as measured data misrepresents both their precision and their provenance.

Copyright and attribution

Digitising a figure does not grant you rights in it. The figure, and often the underlying data, belong to the authors and publisher. Extracting numbers may be fair use or fair dealing for research, and republishing a reconstructed dataset may not be. Ask the authors for the source data first; it is frequently available and always better than anything read off a picture.

Never use digitised values for

Clinical, dosing, toxicological or exposure decisions.
Engineering design values, material properties, or anything load bearing.
Regulatory submissions, safety cases, or certification records.
Meta-analysis or published reanalysis without stating the digitisation method and its uncertainty.
Any claim of measurement precision the source figure does not support.

No warranty. No liability. You assume all risk. This page, its calibration, its extracted values and its exports are provided as-is and as-available, without warranty of any kind, express or implied, including accuracy, completeness, timeliness, reliability, fitness for a particular purpose, merchantability, scientific, engineering or professional suitability, regulatory acceptability, or non-infringement. To the maximum extent permitted by law, the author accepts no liability for any loss, damage, cost or claim arising from use of or reliance on this page, including any conclusion drawn from digitised values.

Independent verification required. Obtain the underlying data from the original authors wherever possible, verify your calibration by re-reading known points on the figure, and state that values were digitised. The source data, the publication and professional judgement control; where this page disagrees with them, this page is wrong.

Consequential damages excluded. To the maximum extent permitted by law, in no event will the author or maintainers be liable for any direct, indirect, incidental, special, exemplary, punitive or consequential damages, including loss of use, data, profits or business interruption, however caused and on any theory of liability, whether in contract, strict liability or tort including negligence, arising out of or in connection with this page, even if advised of the possibility of such damage.

Indemnity. By using this page you agree to indemnify and hold harmless the author and maintainers against any claim, demand, loss or expense arising out of your use of it or of anything derived from it.

Severability. If any part of this notice is held unenforceable, the remainder continues to apply, and the unenforceable part is limited to the least extent necessary rather than struck out.