OCR-D Browser

An extensible viewer for OCR-D mets.xml files

Screenshot
Features
Installation
- Native
  - From source
  - Via pip
- Docker
Usage
- Native GUI
- Docker service
Configuration
- Configuration file locations
- Configuration file syntax

Screenshot

Features

Browse fileGrps and pages, arranging views next to each other for comparison
PageView: Show original or derived page images with PAGE-XML annotations overlay, similar to PageViewer
ImageView: Show original or derived images (AlternativeImage on any level of the structural hierarchy)
ImageView: Show multiple images at once for different pages (horizontally) or different segments (vertically), zooming freely
XmlView: Show raw PAGE-XML with syntax highlighting, open with PageViewer
TextView: Show concatenated PAGE-XML text annotation
DiffView: Show a simple diff comparison between text annotations from different fileGrps
HtmlView: Show rendered HTML comparison from dinglehopper evaluations

Installation

OCR-D Browser requires Python 3.7 or higher.

Native (tested on Ubuntu 18.04/20.04)

The native installation requires GTK 3.

In any case you need a virtual environment with a current pip version (>=20), preferably your existing OCR-D venv:

Create a current pip venv:

sudo apt install python3-pip python3-venv 
python3 -m venv venv
source venv/bin/activate
pip install --upgrade pip setuptools wheel

From source

git clone https://github.com/hnesk/browse-ocrd.git 
cd browse-ocrd
sudo make deps-ubuntu
make install

Via pip

sudo apt install libcairo2-dev libgirepository1.0-dev
pip install browse-ocrd

Docker

If you have installed Docker, you can build OCR-D Browser as a web service:

docker build -t ocrd_browser .

Or use a prebuilt image from Dockerhub:

docker pull hnesk/ocrd_browser

Usage

Native GUI

Start the app with the filesystem path to the METS file of your OCR-D workspace:

browse-ocrd ./path/to/mets.xml

You can still open another METS file from the UI though.

Docker service

When running the webservice, you need to pass a directory DATADIR which (recursively) contains all the workspaces you want to serve. The top entrypoint http://localhost/ will show an index page with a link http://localhost/browse/... for each workspace path. Each link will run browse-ocrd at that workspace in the background, and then redirect your browser to the internal Broadway server, which renders the app in the web browser.

To start up, just do:

docker run -it --rm -v DATADIR:/data -p 8085:8085 -p 8080:8080 ocrd_browser

Configuration

Configuration file locations

At startup the following directories a searched for a config file named ocrd-browser.conf

# directories and their default values under Ubuntu 20.04
GLib.get_system_config_dirs()  # '/etc/xdg/xdg-ubuntu/ocrd-browser.conf', '/etc/xdg/ocrd-browser.conf'
GLib.get_user_config_dir()     # '/home/jk/.config/ocrd-browser.conf'  
os.getcwd()                    # './ocrd-browser.conf'

Configuration file syntax

The ocrd-browser.conf file is an ini-file with the following keys:

[FileGroups]
# Preferred fileGrp names for thumbnail display in the Page Browser 
# Comma separated list of regular expressions
preferredImages = OCR-D-IMG, OCR-D-IMG.*, ORIGINAL

# Each Tool has a section header [Tool XYZ]
# At the moment the only defined tool is "PageViewer"  
[Tool PageViewer]
# shell commandline to execute with placeholders  
commandline = /usr/bin/java -jar /home/jk/bin/JPageViewer/JPageViewer.jar --resolve-dir {workspace.directory} {file.path.absolute}

The commandline string will be used as a python format string with the keyword arguments:

workspace : The current ocrd.Workspace, all properties get shell escaped (by shlex.quote) automatically.
file : The current ocrd_models.OcrdFile, all properties get shell escaped (by shlex.quote) automatically, also there is an additional property path with the properties absolute and relative, so {file.path.absolute} will be replaced by the shell quoted absolute path of the file.

Note: You can get PRImA's PageViewer at Github.

kba / browse-ocrd