Docker usage

Below are examples for using the image-dataset-converter library via its Docker images.

Interactive session

The following command starts an interactive session, mapping the current working directory to /workspace:

docker run --rm -u $(id -u):$(id -g) \
    -v `pwd`:/workspace \
    -it waikatodatamining/image-dataset-converter:latest

Conversion pipeline

The following converts an image classification dataset from the sub-dir format (sub-directory names represent the image classification labels) into the ADAMS format, which stores the label in an associated .report file (Java properties file):

docker run --rm -u $(id -u):$(id -g) \
    -v `pwd`:/workspace \
    -it waikatodatamining/image-dataset-converter:latest \
    idc-convert -l INFO \
      from-subdir-ic \
        -l INFO \
        -i /workspace/input/ \
      to-adams-ic \
        -l INFO \
        -o /workspace/output \
        -c classification

NB: The input and output directories are located below the current working directory (pwd).