Script to Cut Datasets

NOTE: This script is for non-commercial research or educational use only.

Use this script to cut ImageNet, Pascal VOC, and Common Objects in Context(COCO) datasets.

Usage

Download the script cut_dataset.py.

Cut ImageNet or Pascal VOC

In a Python console, run the following command after specifying the parameters:

python C:/Users/Downloads/cut_dataset.py \
--source_archive_dir=<full_path_to_source_archive> \
--output_size=<number_of_images> \
--output_archive_dir=<path_to_output_archive> \
--dataset_type=imagenet
--first_image=<image_number>

This command runs the script with the following arguments:

Parameter	Explanation
--source_archive_dir=<full_path_to_source_archive>	Full path to the downloaded archive including the name
--output_size=<number_of_images>	Number of images to be left in a smaller dataset
--output_archive_dir=<path_to_output_archive>	Full directory to the smaller dataset excluding the name
--dataset_type=<dataset_type>	Type of the source dataset (imagenet or voc)
--first_image=<image_number>	Optional. The number of the image to start cutting from. Specify if you want to split your dataset into train/val subsets. The default is 0.

Cut COCO

In a Python console, run the following command after specifying the parameters:

python C:/Users/Downloads/cut_dataset.py \
--source_images_archive_dir=<full_path_to_source_images_archive> \
--source_annotations_archive_dir=<full_path_to_source_annotations_archive> \
--output_size=<number_of_images> \
--output_archive_dir=<path_to_output_archive> \
--dataset_type=coco
--first_image=<image_number>

This command runs the script with the following arguments:

Parameter	Explanation
--source_images_archive_dir=<full_path_to_source_images_archive>	Full path to the downloaded archive with images, including the name
--source_annotations_archive_dir=<full_path_to_source_annotations_archive>	Full path to the downloaded archive with annotations, including the name
--output_size=<number_of_images>	Number of images to be left in a smaller dataset
--output_archive_dir=<path_to_output_archive>	Full directory to the smaller dataset excluding the name
--dataset_type=<dataset_type>	Type of the source dataset
--first_image=<image_number>	Optional. The number of the image to start cutting from. Specify if you want to split your dataset into train/val subsets. The default is 0.