Yoink is a tool for extracting data from scientific figures.
I'm often faced with this problem: I have a scientific publication with data that I want to use, however the raw data is not available. It is only provided in rasterized images. The figures are of good quality: the line plots provide axes with labels and ticks; the pseudocolor (false-color) images have axes and a colorbar. In theory all the needed information is there, it is just not particularly convenient.
Yoink provides scripts to yoink the data out of rasterized images as well as a set of backend independent matplotlib widgets that assist in doing so.
Yoink does not require exotic dependencies, only python 2.7, numpy, scipy, and matplotlib. (planned) If scikit-image is available, yoink can guess the cropping and rotation of a figure.
The easiest way to use yoink is to use the executables yoink_cmap and yoink_points (planned).
$ yoink_cmap wacky_figure.png -o output.txt