Simple tool to split a multi-label coco annotation dataset with preserving class distributions among train and test sets.

The code is an updated version from akarazniewicz/cocosplit original repo, where the functionality of splitting multi-class data while preserving distributions is added.

Installation

cocosplit requires python 3 and basic set of dependencies:

specifically, in addition to the requirements of the original repo, (scikit-multilearn) is required, it is included the requirements.txt file

pip install -r requirements.txt

Usage

The same as the original repo, with adding an argument (--multi-class) to preserve class distributions The argument is optional to ensure backward compatibility

$ python cocosplit.py -h
usage: cocosplit.py [-h] -s SPLIT [--having-annotations]
                    coco_annotations train test

Splits COCO annotations file into training and test sets.

positional arguments:
  coco_annotations      Path to COCO annotations file.
  train                 Where to store COCO training annotations
  test                  Where to store COCO test annotations

optional arguments:
  -h, --help            show this help message and exit
  -s SPLIT              A percentage of a split; a number in (0, 1)
  --having-annotations  Ignore all images without annotations. Keep only these
                        with at least one annotation
  --multi-class         Split a multi-class dataset while preserving class
                        distributions in train and test sets

Running

$ python cocosplit.py --having-annotations --multi-class -s 0.8 /path/to/your/coco_annotations.json train.json test.json

will split coco_annotation.json into train.json and test.json with ratio 80%/20% respectively. It will skip all images (--having-annotations) without annotations.