dataset-librarian 1.0.0

Dataset librarian is a tool to download and apply the preprocessing needed for the list of supported datasets

Dataset Librarian

Installation Process

python -m pip install dataset-librarian

For more information check the dataset-librarian PyPI package

Datasets

Dataset name	Description	Download	Preprocessing	command
brca	Breast Cancer dataset that contains categorized contrast enhanced mammography data and radiologists’ notes.	supported	A prerequisite: Use a browser, download the Low Energy and Subtracted images, then provide the path to the directory that contains the downloaded images using --directory argument.	python -m dataset_librarian.dataset -n brca --download --preprocess -d <path to the dataset directory>
tabformer	Credit card data for TabFormer	supported	not supported	python -m dataset_librarian.dataset -n tabformer --download
dureader-vis	DuReader-vis for document automation. Chinese Open-domain Document Visual Question Answering (Open-Domain DocVQA) dataset, containing about 15K question-answering pairs and 158K document images from the Baidu search engine.	supported	not supported	python -m dataset_librarian.dataset -n dureader-vis --download
msmarco	MS MARCO is a collection of datasets focused on deep learning in search	supported	not supported	python -m dataset_librarian.dataset -n msmarco --download
mvtec-ad	MVTEC Anomaly Detection DATASET for industrial inspection. It contains over 5000 high-resolution images divided into fifteen different object and texture categories.	supported	supported	python -m dataset_librarian.dataset -n mvtec-ad --download --preprocess -d <path to the dataset directory>

Command-line Interface

Input Arguments	Description
--list (-l)	list the supported datasets.
--name (-n)	dataset name
--directory (-d)	directory location where the raw dataset will be saved on your system. It's also where the preprocessed dataset files will be written. If not set, a directory with the dataset name will be created.
--download	download the dataset specified.
--preprocess	preprocess the dataset if supported.

Python API

from dataset_librarian.dataset_api.download import download_dataset
from dataset_librarian.dataset_api.preprocess import preprocess_dataset

# Download the datasets
download_dataset('brca', <path to the raw dataset directory>)

# Preprocess the datasets
preprocess_dataset('brca', <path to the raw dataset directory>)

Building for source

Clone the Model Zoo for Intel® Architecture repository and navigate to the dataset_api directory.

git clone https://github.com/IntelAI/models.git
cd models/datasets/dataset_api
python -m pip install --upgrade pip build setuptools wheel
python -m pip install .