Open fattynoparents opened 1 month ago
Did you try training a baseline labeling model? Maybe you solve this issue with your own model.
If not, is the command ketos segtrain
Did you try training a baseline labeling model? Maybe you solve this issue with your own model.
Yes this is my trained models, both baseline- and regionmodel. I finetuned them separately with --suppress-regions
and --suppress-baselines
parameters.
I'm running kraken segmentation inference with the following command, including two different models for baselines- and regions- recognition:
There arise some cases when a baseline goes too close the border to a region and then the kraken engine fails to assign it to this region. In eScriptorium it looks like the baseline is assigned to the rose region: However if I look closer, the baseline is not assigned and cannot be assigned even manually:
Here are also the corresponding parts from XML files that I get from kraken. The dummy region to which the above mentioned baseline is assigned:
and the empty Författare region to which the baseline should ACTUALLY be assigned:
Is it something that can be fixed somehow? Thanks in advance.