Skip to content
Snippets Groups Projects
Select Git revision
  • bump-griffe
  • bump-boto3
  • bump-mlflow-skinny
  • half_precision
  • bump-mkdocstrings-python
  • bump-mkdocs-material
  • bump-numpy
  • main default protected
  • bump-wandb
  • unit-test-polygon-bug
  • bump-albumentations
  • bump-matplotlib
  • bump-scipy
  • bump-tensorboard
  • bump-lxml
  • bump-torchaudio
  • bump-torchvision
  • bump-torch
  • bump-black
  • bump-imageio
  • 0.2.2b2
  • 0.2.2b1
  • 0.2.2a2
  • 0.2.2a1
  • 0.2.1
  • 0.2.0
  • 0.2.0rc12
  • 0.2.0rc11
  • 0.2.0rc10
  • 0.2.0rc9
  • 0.2.0rc8
  • 0.2.0rc7
  • 0.2.0rc6
  • 0.2.0rc5
  • 0.2.0rc4
  • 0.2.0rc3
  • 0.2.0rc2
  • 0.2.0rc1
  • 0.2.0-dev6
  • 0.2.0-dev5
40 results
Compare
Name Last commit Last update
..
augmentation.md
index.md
jeanzay.md
parameters.md
index.md

Train

Use the teklia-dan train document command to train a new DAN model. It is able to train a DAN model at line or document-level and evaluate it.

Examples

Document

To train DAN on documents:

  1. Set your training configuration in dan/ocr/document/train.py. Refer to the dedicated section for a description of parameters.
  2. Run teklia-dan train document.
  3. Look into evaluation results in the output folder:
    • checkpoints contains model weights for the last trained epoch and for the epoch giving the best valid CER.
    • results contains the tensorboard log file, the parameters file, and the evaluation results for the best epoch.

Line

To train DAN on lines, run teklia-dan train document with a line dataset.

Additional page