How to extract data from a graph in a PDF
A graph in a PDF is usually stored as a drawing: lines and shapes with exact positions. That means the data can be read back exactly, without guessing from pixels.
How it works
When a PDF holds the chart as a drawing (most figures in papers do), the values are read from the drawing itself. Nothing is estimated from pixels, so a point is as precise as the file is.
The tick labels give the axis scale, the legend gives the series names, and each bar, marker and line corner gives a value. Error bars are read at both ends.
Steps
- Upload the PDF. There is no need to cut the figure out first.
- Each chart found comes back on its own tab. Pick pages if the paper is long.
- Download one Excel file with a sheet per chart.
When the PDF does not help
Some PDFs hold the figure as a picture (older scans, some journal archives). Then the page is read as a picture instead, and you should check the values on the chart.
Charts drawn with matplotlib, R (ggplot2 and base graphics) and gnuplot are what we test on. A chart from another program may be read in part; the result says how many shapes in the plot were not read.
Questions
Is the PDF kept?
No. It is read and then discarded.
Does reading a PDF use AI?
No. Charts drawn in a PDF are read by ordinary code on our server and never go to an AI model.