We package all the layout annotation service (the annotation interface and active learning modeling server) inside docker containers. The installation process is very straightforward and simple:
git clone git@github.com:Layout-Parser/annotation-service.git
cd annotation-service
DATA=./data CONFIG=labeling-config.xml MODEL=model.py docker-compose up --build -d
labeled
. In the 3rd command, the environmental variables DATA
, CONFIG
, and MODEL
are used to set the labeling data directory, Label Studio configuration file, and ML backend model file, respectively.
DATA
is for the folder containing all the images for labeling. By default, DATA=./data
. CONFIG
is the configuration file for initializing the label-studio interface. The default value is CONFIG=horizontal-layout.xml
, and you could find more examples in labeling/configs
.MODEL
is for the script that generates the model prediction. The default value is MODEL=model.py
.