The marquee corners definition in the Photo Tagging gramplet is defined using an 0٪‐100٪ integer percentage. (Which allows the same rectangle definition to be used with different resolution copies of the same media object.)
That’s ok for low resolution Media objects. But it is hard to generate accurate thumbnails with high resolution media objects with lots of marquee rectangles. (For example, a family reunion group photo.)
As an example, I wanted to use a grid of flags (in SVG format) as a media object. Then make thumbnails for the galleries of the various countries. But the Photo Tagging’ rectangular marquees did not have enough resolution. (It also doesn’t like Places for its tagging.)
I don’t know if what I’m going to write is related to this topic or if it should be the subject of another question:
Indeed, for some high definition images, for example a digitized registry page downloaded from FamilySearch, it is impossible to zoom in deep enough into the image to correctly select what you are trying to show, for example a signature at the bottom of a document on an image containing several, and for the document itself, if selecting it is quite simple, it is often impossible to zoom in deep enough into the image to display it in its entirety and succeed in reading its content directly into Gramps media window, while opening the image in external software allows you to zoom in and see this content correctly
That could be another side of the same problem for the Photo Tagging gramplet. I have not tested it there.
But I think you are reporting this “insufficent zoom” behavior in the core’s Create a reference region for the Media Reference editor from the Gallery tab, not a 3rd-party add-on. Is that correct?
If so, it is definitely a different bug report… but still fits this discussion.
