README.md

Note: If you're looking for a GUI tool to help with your Machine Learning data annotation task, please have a look at LabelImg and CVAT. ImageLabeler (or rather ImageCategorizer coined by user @de-ar) does not export to standard annotation formats.

ImageLabeler

GUI tool to label images. Labels are appended at the front of the image filename. The images are taken from a source directory and the labeled images are placed in the destination directory.

Installing dependencies

$ pip install pipenv
$ pipenv install

Configure

LABEL_CLASSES in ImageLabeler.py can be updated to configure the labeler. This is a list, with class names. The positions of the class names matter, as the buttons and labels will follow this positioning.

Run

usage: ImageLabeler.py [-h] [-b BACKUP_DIRECTORY]
                       source_directory destination_directory

Run this app to label images by appending the label at the front of the
filename.

positional arguments:
  source_directory      Directory from where images will we be
                        taken.Subdirectories included.
  destination_directory
                        Directory where labeled images will be stored.

optional arguments:
  -h, --help            show this help message and exit
  -b BACKUP_DIRECTORY, --backup-directory BACKUP_DIRECTORY
                        Directory to store images that were left unlabeled.
                        Leaving this option willadd a label to the image as
                        per the label formatter used in code.

Example

$ python ImageLabeler.py src_dir des_dir -b unlabeled_dir

Demo

Clone the repo. Run the demo using the following command.

$ python ImageLabeler.py ./Demo/unsorted ./Demo/sorted -b ./Demo/unlabeled

LabelFormatter

You can define your own LabelFormatter. Check out LabelUtilities.py.