We compare our auto-digitizer to the current available “plot digitizer”. In the plot digitizer tool, users must to click on each data point they want to extract. Increasing with the number of data points there are in the graph, this is tedious and timeconsuming work. In our auto-digitizer, if the OCR can detect the axis label numbers successfully, the total process time is a constant of only 20s and requires less effort.
Our current version of the auto-digitizer is still sensitive to the format of input graphs and exhibits a slight offset in extracted data values to the original graphs. Moving forward, we will focus improving the robustness to graph format and background noise, and develop a methodology for data mapping that yields more accurate results.