Skip to content

Latest commit

 

History

History
109 lines (75 loc) · 3.61 KB

Readme.md

File metadata and controls

109 lines (75 loc) · 3.61 KB
Raw

API

See the related documentation in the API page.

Running the server

Running the server with docker

See the related documentation to run using docker.

Running the server from sources

The simplest way to get started is to:

  1. Build the C++ code as explained in the build instructions
  2. Copy build/back/c++-bindings/soducocxx.cpython-38-x86_64-linux-gnu.so in back/soducocxx.cpython-38-x86_64-linux-gnu.so
  3. Create a settings config file and export its path: export SODUCO_SETTINGS="<PATH_TO_SETTINGS.CFG>"
  4. Run the server with python -m flask run

Fast build and test using Docker for development

  1. Make sure your configuration for the frontend is correct: update the port in ~/.config/soduco_annot/general_config.ini so that port=8000
  2. Configure docker/.env to allow direct connection to the server (create if needed from any docker/.env-sample* files):
    • update PATH_DIRECTORIES to you local path
    • set REVERSE_PORT=8000
    • remove any line with SCRIPT_NAME
  3. Use the following dockerfile to build the core library:
    • download this file to an empty directory
    • build the docker image for the builder with docker build -t soduco/back_builder -f docker/soduco-back-cppbuilder.dockerfile .
    • you can now run the builder in the root of the backend project: docker run --rm -it -v $(pwd):/app/ soduco/back_builder sh build.sh
    • the artifacts are now at the right place for the deployment image to be built
  4. Download (and check for newer) assets
    • sftp soducoshared@geohistoricaldata.org:./public/fr_ner_directories-20210720.tar.gz
    • wget https://www.lrde.epita.fr/~jchazalo/SHARE/pero_eu_cz_print_newspapers_2020-10-09.tar.gz
    • wget https://download.pytorch.org/models/vgg16-397923af.pth
  5. Build and launch the deployment image using docker-compose (use pipx to install it if needed):
    • cd docker
    • docker-compose up --build
    • the server should run and listen to *:8000

Building from sources

To build the project from C++ sources, see the related documentation

Python Commandline interface

The command line interface is described here

DOM & JSON Format description (outdated)

Document Model

Page      := ( Section_1 | Title_1 )*
Section_1 := ( Column_1 )*
Column_1  := ( Section_2 )*
Section_2 := ( Column_2 )*
Column_2  := ( Entry )*
Entry     := ( Line )*

Terminaux:
Title_1   := 
Title_2   := 
Line      :=

JSON Schema

  • Non-terminal nodes map to JSON Element objects

    {
  "type" : STRING // One of ( "PAGE" | "SECTION" | "COLUMN" | "ENTRY" | "LINE" | "TITLE" )   
  "bbox" : ARRAY OF INT // Element Bounding box coordinates [x, y, width, height]
  "children" : ARRAY OF OBJECTS // List of Children Elements   
}

note that bounding box coordinates are for the deskewed image

  • Text nodes (TITLE and LINE) have also the "text" : STRING attribute.

Example

Input Output JSON
JSON file

Deploy to LRDE's public server

Connect to docker.lrde.epita.fr

cd /etc/docker/compose/directory-annotator

# logs
docker-compose -p directory-annotator logs backend

# Up
systemctl start docker-compose@directory-annotator.service

# Stop
docker-compose -p directory-annotator stop