Text overlaps on some ALTO files

[benwbrum]

We'e run into a handful of ALTO files in which the text overlay font is too large, resulting in overlapping text regions and unreadable text.

Here's an example:

• Configure OSD with
  tileSources: ["{\"type\":\"image\",\"url\":\"https://fromthepage.com/images/uploaded/32108542/MS0236-S01-004_003-0.jpg\"}"
• Configure the text layer with
  textlayer.loadOCR('https://fromthepage.com/nal/charles-c-plitt-collection/ms0236-s01-004/34013749/alto_xml')

Results:

[rsimon]

Hi @benwbrum,

Would it be possible to post the ALTO file here? I can access the image, but the ALTO seems behind a login. (Otherwise, coud you email it to me?)

A simple fix would (probably) be to reduce the global scaling factor that's applied by the overlay plugin. That's now hard-coded in the code. I could make it configurable from the outside. Although that wouldn't be much help, I guess. Are the other pages coming out fine, more or less? In this case, reducing the global scaling factor would likely make text on other pages too small.

Either way, tweaking the scaling factor will always be a bit poking at the problem with a stick. Not sure how to make this smarter & more sustainable... One solution I could think of would be, perhaps, to compare the actual size of the rendered text with the size of the annotation (which isn't rendered - but known internally). I can do some testing along these lines. Do let me know if you have any further ideas though!

[benwbrum]

Absolutely! Github won't let me add XML files to issues, so I've posted it as a Gist which you may be able to point Annotorious at.

Regarding other pages, we see fonts that are too big more often than we don't.

I was wondering if it would be possible to detect overlapping text, then downscale if any was present. I don't know enough about whether that's visible to you -- presumably the annotations themselves are not overlapping, and the text just over-runs their bounding boxes? (This may be what you mean on by comparing actual size with annotation size.)

Another option would be to add a font scaling selector to the UI, so that end users can change the font size.

[rsimon]

Thanks!

Ah, that's weird. I'm getting a good result with the current version of the plugin:

Do you know which version you are using? Maybe I got something mixed up with the last PR, and you're still on the old version with Less Smartness™. I'll rebuild the bundle and send another PR. Let's see if this fixes the problem.

[benwbrum]

That sounds like a great plan. There were some issues with strange commits in the last two PRs, so it's very possible that I messed something up when merging your changes in.