Robert Sachunsky

model behaviour on text vs. non-text

1

If training was done on [DIBCO datasets](https://github.com/ajgallego/document-image-binarization#datasets), some of which according to @VChristlein are representing non-text differently, how can the models here be expected to perform on non-textual parts (separators,...

Depends on OCR-D/core#1240 (i.e. `ocrd>=3.0`).

In the navigation view on the left where the thumbnail images are displayed an pages can be selected, pages can only be identified via their file ID (and the contiguous...

The current [legacy](https://packaging.python.org/en/latest/guides/packaging-namespace-packages/) approach to namespace packaging does not work with development/editable installation. See https://github.com/OCR-D/ocrd_all/issues/433 This PR fixes it for this package – analogous PRs to the other qurator repos...

attempt at migration towards https://github.com/OCR-D/core/pull/1240 no tests yet, I will try to add some CI here @kba see `ocrd_segment.replace_original` for the use case of passing `PageType` in `OcrdPageResultImage.images.alternative_image`. – It...

just a very superficial update – we need to look more thoroughly, but the current information on the website is highly misleading, so this is urgent.

confidence output

4

It would be great if Eynollah would annotate the confidence score of detected segments. From PAGE, this naturally can be represented via `Coords/@conf`. Of course, it would necessitate another refactoring,...

ocrd_utils.initLogging: also add handler to root logger

7

As a follow-up to https://github.com/OCR-D/core/commit/ccb416b13e7f91781568fda8e60ad8182bfea88c, but more general: `initLogging` should _always_ affect the root logger `''`, too. (Not just when disabling logging.) Reasons: - to be consistent with file config...

see https://github.com/OCR-D/core/blob/81c0b4c01eb1db54999c5db8feae120df2cc1a7b/src/ocrd_network/cli/client.py#L91 for ideas, see [here](https://github.com/OCR-D/core/issues/1032#issuecomment-1511482113), so in this case `ocrd network client workflow list|download|upload|run` where the latter should look like `ocrd process` (i.e. block) As an extra, we could...