Dataset Preparation#

Here we show how to download datasets that integrate with SportsLabKit.

SoccerTrack#

The SoccerTrack Dataset is available to download from Kaggle.

You will need a Kaggle account to download the dataset!

How to download from Kaggle using KaggleDownloader#

SoccerTrack a helper class called KaggleDownloader to download the dataset from directly from Kaggle. However, you will need to authenticate your kaggle ccount first. To do so, follow the intructions by the KaggleDownloader class at initialisation.

import sportslabkit as slk

dl = slk.datasets.KaggleDownloader()

authenticate:0172  💬| Authentication successful.

Without a Kaggle account you will not be able to download the dataset with this method. After authenticating your Kaggle account, you can download the dataset by calling the download() method. The dataset will be downloaded to the directory specified by the path argument. The default value is `/path/to/soccertrack/soccertrack/datasets/’.

%%script false --no-raise-error # comment out this line to run this cell

dl = KaggleDownloader()

path = dl.download(force=False)

It is also possible to download files individually by specifying the file argument. The default value is None, which will download all files. The file must be a string representing the name of the file/folder to download. For example, to download the top-view dataset, you can call download(file='topview').

%%script false --no-raise-error # comment out this line to run this cell
from sportslabkit.logger import show_df # This just makes the df viewable in the notebook.

save_path = dl.download(file_name='wide_view/annotations/F_20200220_1_0000_0030.csv', path='.', force=True)
df = slk.load_df(save_path)

show_df(df.head())

The get_path method returns the path to the downloaded dataset. A usage example is shown below.

# show the available datasets
slk.datasets.get_path()

# show the path to the dataset
wide_view_path = slk.datasets.get_path('wide_view')

# get a csv-mp4 pair from the wide-view dataset
wide_view_csv = slk.datasets.get_path('wide_view/videos/F_20200220_1_0000_0030.csv')
wide_view_mp4 = slk.datasets.get_path('wide_view/annotations/F_20200220_1_0000_0030.mp4')

Available keys:
 - GNSS
 - top_view
 - wide_view
 - drone_keypoints
 - fisheye_keypoints
 - gnss_keypoints

If the path you specify is invalid, the get_path() method will raise a ValueError.

wide_view_path = slk.datasets.get_path('imagenet')

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Cell In[5], line 1
----> 1 wide_view_path = slk.datasets.get_path('imagenet')

File ~/Github/SoccerTrack/sportslabkit/datasets/__init__.py:62, in get_path(dataset)
     60 msg = f"The dataset '{dataset}' is not available. "
     61 msg += f"Available datasets are {', '.join(available)}"
---> 62 raise ValueError(msg)

ValueError: The dataset 'imagenet' is not available. Available datasets are GNSS, top_view, wide_view, drone_keypoints, fisheye_keypoints, gnss_keypoints