Thank you for the link, Zdenko. Since I also want to get the features from ready-to-use traineddata files, that don't come with box/tr files, I started using the methods in tessdatamanager.h and intprotos.h.
Do you know how the features in tr files differ from the data in the intproto section of a traineddata file? I know that intproto mainly contains the 4D (x, y, angle, length) information.
There are four sections for each box in a tr file: mf, cn, if and tb. if seem to be the 3D int features (x, y, angle), but what do the features at mf, cn and tb mean?
Paul