Digitizing Plots with Confidence Intervals from Literature to Inform Model Development #220
Description
Is your feature request related to a problem? Please describe.
A common part of model validation is extracting information from images of plotted data from the literature. This is often done manually and is very time-consuming. It would be helpful for AI tools to assist with digitizing existing literature plots beyond simple x,y coordinates.
Describe the solution you'd like
I am interested in automated digitization of plots where x,y values are separated from confidence intervals in the form of a table with the following columns: x value, y value, lower confidence bound, upper confidence bound.
Describe alternatives you've considered
I have used manual digitization such as Matlab's Digitize function and colleagues use similar approaches where they need to select points manually and come up with their own way to store confidence interval bounds. One tool that uses machine vision to identify points is automeris, but similar tools do not identify confidence intervals.
Additional context (optional)
I am aware that OCR is currently used to exact information from model diagrams. Perhaps that can be used to extract axis information such as title, units, scale (linear/log). Paired with x and y coordinates, this can be used to provide more data for modeling as well as validation. Validation with manual digitizations can prove helpful for model training.