search

grp-bork / formHTR

Handprint text recognition in form documents.
MIT License
1 stars 1 forks source link

Template annotation: textual inputs #22

Open xtrojak opened 9 months ago

xtrojak commented 9 months ago

The textual inputs (variable name corresponding to a ROI, expected value in a residual, path to a dictionary of allowed values, ...) can be handled from command line, but it's not very robust requires changing context between the image and command line). This should be improved.

xtrojak commented 8 months ago

Another way around this is to assing just integer IDs to individual regions and then assign them variable names e.g. via a mapping from text file.