As an alternative to using human labellers to create polygons representing subfigures within a complex scientific illustration (for example by processing images through LabelStudio as outlined in issue #3), investigate using an automated image segmentation approach.
[ ] Gather a set of exemplar images (e.g. those including scale bars, those with completely distinct sub-figures separated by whitespace, those with overlapping subfigures etc) and document these on a new page in the project wiki
[ ] Add a segmentation script segment-illustration.py which accepts each illustration image as a command line argument and creates a set of images representing the subfigures. These should be stored in the directory specified as the second command line argument.
[ ] Add a step to the Makefile which calls this segmentation script for each image available in the data directory
As an alternative to using human labellers to create polygons representing subfigures within a complex scientific illustration (for example by processing images through LabelStudio as outlined in issue #3), investigate using an automated image segmentation approach.
segment-illustration.pywhich accepts each illustration image as a command line argument and creates a set of images representing the subfigures. These should be stored in the directory specified as the second command line argument.Makefilewhich calls this segmentation script for each image available in the data directory