bertsky / workflow-configuration

a makefilization for OCR-D workflows, with configuration examples
Apache License 2.0
9 stars 4 forks source link

workflow-configuration not running #4

Closed EEngl52 closed 4 years ago

EEngl52 commented 4 years ago

I adapted gt-binarize-page-olena-sauvola-clip-resegment-dewarp-ocr-ocropy-tesseract.mk by renaming the input file group and exchanging the first processor olena-binarize with cis-ocropy-binarize (see attachment).

When I try to run make -r test.mk /test/data/helmstedt or -f it would just tell me that there's nothing to do in this workspace. The same holds true when using a relative path to the workspace. -s doesn't return any message.

test.zip

bertsky commented 4 years ago

Looking into it now.

Just 2 things ahead:

bertsky commented 4 years ago

Okay, at that point – without your input data – I need your help. Can you please show the result of running LC_MESSAGES=C make -d -f test.mk helmstedt?

BTW, this configuration is not going to work anyway:

$(BIN): TOOL = ocrd-cis-ocropy-binarize
$(BIN): PARAMS = "level-of-operation": "line"

CLIP = $(BIN)-CLIP

$(CLIP): $(BIN)
$(CLIP): TOOL = ocrd-cis-ocropy-clip

...because you cannot clip segments that already have AlternativeImage on them – this makes them incomparable. (Derived images one level above the level-of-operation are fine of course.) So your binarization needs to use PARAMS = "level-of-operation": "page".

EEngl52 commented 4 years ago

ah, ok, thanks for your explanations

there's the output from your command. I could also zip my workspace and upload it if that helps output.txt

bertsky commented 4 years ago

Are you sure you ran with helmstedt not with the absolute path?

Sorry, I know this should be irrelevant. Just trying to understand.

Also, yes, if it's possible, please upload your workspace (the mets.xml and empty directories will sufffice though).

EEngl52 commented 4 years ago

sorry, used the absolute path. there's the output for helmstedt

erroutput.txt

and there's my workspace. I had to delete the original images in OCR-D-IMG as they were too big to upload

helmstedt.zip

EEngl52 commented 4 years ago

just changed level of operation to page and ran it on a new workspace with just OCR-D-IMG as input. This also produced the same error message

bertsky commented 4 years ago

I think I know what the issue is. This happens when you don't install (as documented by the README). However, I am thinking of a solution without installation anyway...

bertsky commented 4 years ago

... the current master should allow that usage now (and also complain when no workspaces can be found). @EEngl52 could you please try it out, so we can close?

EEngl52 commented 4 years ago

I just updated and tried the new version. It works perfectly fine, also with relative paths. thx @bertsky !!