Introduction

The repository adds simple class for adding a wrapper on top of kaggle cli tools for ease of downloading competition datasets, listing competitions, downloading and un-zipping datasets.

It aims to make it easy to use kaggle datasets directly from python scripts. It's main usecase is to be used as a submodule in other repos/ or used as a standalone package. If the library is used as a standalone package - the data is downloaded and stored in data folder. In case of using the package as a submodule - the data is downloaded in a data folder created in the base folder of the calling script.

Installation

The repo relies on setuptools and kaggle library packages. To install the repo please use one of the following options:

pip install -U kaggle setuptools
python setup.py install

It also assumes that presence of Kaggle User ID and public API credentials at

On Windows: C:\Users\<Windows-username>\.kaggle\kaggle.json

Others: ~/. kaggle/kaggle.json

After the above steps: the repo can be installed from git using:

pip install git+https://github.com/sunny1401/kaggle_utils.git#egg=kaggle-cli-wrapper

The project can also be installed using pip:

pip install kaggle-cli-wrapper

Code details

DataAPI class

To list all available datasets

from kaggle_cli_wrapper import KaggleDataApi

kda = KaggleDataApi(call_path=__file__)
kda.list_all_kaggle_datasets(search_term="cityscapes")

The call_path argument is required to decide the folder where downloaded files are stored.

The function would return all available datasets with cityscape. This would also be saved as a .txt file.

Currently the list is sorted by "votes". To change this sorting please use the argument sort_by. Allowed values of sort by are:

'hottest', 'votes', 'updated', 'active', 'published'

To list all available competition datasets

from kaggle_cli_wrapper import KaggleDataApi

kda = KaggleDataApi(call_path=__file__)
kda.list_all_kaggle_competitions(search_term="cityscapes")

Currently the list is sorted by "earliestDeadline". To change this sorting please use the argument sort_by. Allowed values of sort by are:

'grouped', 'prize', 'earliestDeadline','latestDeadline', 'numberOfTeams', 'recentlyCreated'

To download datasets

from kaggle_cli_wrapper import KaggleDataApi

kda = KaggleDataApi(call_path=__file__)
kda.download_kaggle_dataset(dataset_name="cityscapes_train_val_test", is_competition_dataset=False)

To download a competition dataset, is_competition_dataset needs to be set to True

ScoringAPI

The code currently supports minimum functionality for submission of files to a competition and getting scores for the competition

To submit to a competition

from kaggle_utils.kaggle_cli_wrapper import KaggleScoringsApi

kaggle_scoring_api = KaggleScoringsApi(competition_name="facial-keypoints-detection")

kaggle_scoring_api.submit_solution(submissions_file=submission_path, description="facial_keypoint_vanilla_cnn")
kaggle_scoring_api.get_top_scores()

Remaining Issues

• improve error handling
• allow for saving of stdout