README.MD

OCR-D workflow for Taverna

To develop or edit the workflows it's recommended to use Taverna Workbench Core 2.5 (see requirements). For executing workflow the taverna commandline tool is sufficient.

Requirements

• Java 7 or 8 (Ubuntu: 'sudo apt-get install openjdk-8-jdk') (will at least not work with java 11)
• Git
• Processors (check GitHub to see how they are created)

Installation

Installation using Docker

The OCR-D workflow and all processors may be also exectued inside a docker container. Please refer to OCR-D workflow for Taverna using Docker

Local Installation

Install Processors

First all needed processors should be installed. Please refer to GitHub

Checkout Taverna OCR-D Workflow via GitHub

user@localhost:~$mkdir git
user@localhost:~$cd git
user@localhost:~/git$git clone https://github.com/OCR-D/taverna_workflow.git
user@localhost:~/git$cd taverna_workflow/
user@localhost:~/git/taverna_workflow$

Create Directory for all configuration files and documents

Create a directory where you can store all documents. There should be at least 500 MB of free disc space left. After that install taverna workflow into the created directory.

user@localhost:~$mkdir -p ~/ocrd/taverna
user@localhost:~$cd ~/git/taverna_workflow/
user@localhost:~/git/taverna$bash installTaverna.sh ~/ocrd/taverna

First Test of OCR-D Workflow

To check if the installation works fine you can start a first test.

user@localhost:~$cd ~/ocrd/taverna
user@localhost:~/ocrd/taverna$bash startWorkflow.sh parameters.txt
[...]
Outputs will be saved to the directory: /.../Execute_OCR_D_workfl_output
# The processed workspace should look like this:
user@localhost:~/ocrd/taverna$ls -1 workspace/example/data/
metadata
mets.xml
OCR-D-GT-SEG-BLOCK
OCR-D-GT-SEG-PAGE
OCR-D-IMG
OCR-D-IMG-BIN
OCR-D-IMG-BIN-OCROPY
OCR-D-OCR-CALAMARI_GT4HIST
OCR-D-OCR-TESSEROCR-BOTH
OCR-D-OCR-TESSEROCR-FRAKTUR
OCR-D-OCR-TESSEROCR-GT4HISTOCR
OCR-D-SEG-LINE
OCR-D-SEG-REGION

Each sub folder starting with 'OCR-D-OCR' should now contain 4 files with the detected full text.

The metadata sub directory

The subdirectory 'metadata' contains the provenance of the workflow all intermediate mets files and the stdout and stderr output of all executed processors.

Test your own Documents

If you want to test this workflow with your own documents you have to pass the workspace directory with a mets.xml inside to the shell script.

user@localhost:~$cd ~/ocrd/taverna
user@localhost:~/ocrd/taverna$bash startWorkflow.sh parameters_all.txt /path/to/workspace/containing/mets

Use predefined Workflow

Taverna comes with 2 predefined workflows.

1. parameters_best.txt: Workflow expected to produce best results
2. parameters_fast.txt: Workflow expected to produce good results even on slower machines.

user@localhost:~$cd ~/ocrd/taverna
user@localhost:~/ocrd/taverna$bash startWorkflow.sh parameters_best.txt /path/to/workspace/containing/mets

Configure your own Workflow

Before a workflow can be started, it must be configured first. You may use the predefined workflows as a starting point.

# Make a copy of a parameters and a workflow configuration
user@localhost:~/ocrd/taverna$cp conf/parameters_all.txt conf/my_parameters.txt
user@localhost:~/ocrd/taverna$cp conf/workflow_configuration_all.txt conf/my_workflow_configuration.txt

Now adapt the configuration files due to your needs. (See examples inside the files)

ℹ️ Make sure that all processors used by the workflow are installed.

Test Processors

For a fast test if a processor is available try the following command:

# Test if processor is installed e.g. ocrd-cis-ocropy-binarize
user@localhost:~/ocrd/taverna$ocrd-cis-ocropy-binarize -J
{
 "executable": "ocrd-cis-ocropy-binarize",
 "categories": [
  "Image preprocessing"
 ],
 "steps": [
  "preprocessing/optimization/binarization",
  "preprocessing/optimization/grayscale_normalization",
  "preprocessing/optimization/deskewing"
 ],
 "input_file_grp": [
  "OCR-D-IMG",
  "OCR-D-SEG-BLOCK",
  "OCR-D-SEG-LINE"
 ],
 "output_file_grp": [
  "OCR-D-IMG-BIN",
  "OCR-D-SEG-BLOCK",
  "OCR-D-SEG-LINE"
 ],
 "description": "Binarize (and optionally deskew/despeckle) pages / regions / lines with ocropy",
 "parameters": {
  "method": {
   "type": "string",
   "enum": [
    "none",
    "global",
    "otsu",
    "gauss-otsu",
    "ocropy"
   ],
   "description": "binarization method to use (only ocropy will include deskewing)",
   "default": "ocropy"
  },
  "grayscale": {
   "type": "boolean",
   "description": "for the ocropy method, produce grayscale-normalized instead of thresholded image",
   "default": false
  },
  "maxskew": {
   "type": "number",
   "description": "modulus of maximum skewing angle to detect (larger will be slower, 0 will deactivate deskewing)",
   "default": 0.0
  },
  "noise_maxsize": {
   "type": "number",
   "description": "maximum pixel number for connected components to regard as noise (0 will deactivate denoising)",
   "default": 0
  },
  "level-of-operation": {
   "type": "string",
   "enum": [
    "page",
    "region",
    "line"
   ],
   "description": "PAGE XML hierarchy level granularity to annotate images for",
   "default": "page"
  }
 }
}
user@localhost:~/ocrd/taverna$

Execute your own Workflow

If workflow is configured it can be started.

user@localhost:~/ocrd/taverna$bash startWorkflow.sh my_parameters.txt /path/to/workspace/containing/mets

More Information

• Workflow for OCR-D
• Taverna
• Taverna Commandline Tool