New mode to split prediction by starting tokens

Closes #198 (closed)

added P1 label

assigned to @mblanco

changed title from Split prediction by ner to New mode to split prediction by starting tokens

I got these results

with this command:

cd tests/data/prediction
teklia-dan predict \
  --image images/2c242f5c-e979-43c4-b6f2-a6d4815b651d.png \
  --model popp_line_model.pt \
  --parameters parameters.yml \
  --charset charset.pkl \
  --tokens tokens.yml \
  --output prediction/ \
  --attention-map \
  --attention-map-level ner \
  --predict-objects \
  --threshold-method otsu \
  --confidence-score \
  --confidence-score-level ner

requested review from @yschneider

added 1 commit

• cd44ce7a - Fix documentation

Compare with previous version

resolved all threads

approved this merge request

added 1 commit

• 1a33414f - Remove Level.list()

Compare with previous version

resolved all threads

added 7 commits

• 1a33414f...29b43932 - 3 commits from branch main
• 950670eb - New mode to split prediction by starting tokens
• e79fbc84 - Fix documentation and handle errors
• 5fbcd90e - Fix documentation
• 79ce0bd7 - Remove Level.list()

Compare with previous version

enabled an automatic merge when the pipeline for 79ce0bd7 succeeds

merged