Automatically exported from
Latest commit a035faa Mar 17, 2015 @gtoderici Update README


The YouTube Sports-1M Dataset

We provide links to 1,133,158 YouTube videos annotated with 487 sports labels.
The annotations were generated automatically using the the YouTube Topics,
which has a public API accessible at:

Compression: in order to save space, we compressed all the text files referenced
below via gzip. To decompress run gzip -d <filename.gz> in order to obtain the
original filename that we refer to below.

The files included in this package are:

original/test_partition.txt - this contains the testing partition.
original/train_partition.txt - this contains the training partition.

The format for the training/testing partitions is as follows:
URL<space><CSV of Label Indices>

For example, the following line is a valid input: 168,169

This assigns labels 168, and 169 to the video found at given URL.

labels.txt - this file contains the human-readable labels for the train/test
partitions. The first line in the file is assumed to correspond to label 0
(boomerang), and the last corresponds to index 486 (model aircraft).

sports_mids.txt - this file  contains the Machine IDs necessary to retrieve
videos from via the topics search API. Each line contains the
human-readable class name, and the YouTube topic ID, which may be used to
directly retrieve videos for the given class using the API below:

Extra files:
cross-validation/all_urls.txt - all URLs and labels bundled together (good
starting point if you want to make cross-validation partitions). The format is
as explained above.

cross-validation/sportsX_train.txt & cross-validation/sportsX_test.txt for X
having values from 0 to 9. These are partitions for 10-fold cross-validation.
Since a video may have more than one label, it may appear both in training and
in testing. For example, video ABPsSSS2uY0 appears in fold 0 with class 49 for
training and class 26 for testing.

Additional Information

Wiki page:

  title     = {Large-scale Video Classification with Convolutional Neural Networks},
  author    = {Andrej Karpathy and George Toderici and Sanketh Shetty and Thomas Leung and Rahul Sukthankar and Li Fei-Fei},
  year      = {2014},
  booktitle = {CVPR}

Supplemental materials:


This data set is made available under a Creative Commons License:

Attribution 3.0 Unported (CC BY 3.0) Human-Readable Summary

You are free:
 - to Share -- to copy, distribute and transmit the work;
 - to Remix -- to adapt the work;
and to make commercial use of the work.

Under the following conditions:

* Attribution -- You must attribute the work in the manner specified
by the author or licensor (but not in any way that suggests that
they endorse you or your use of the work).

With the understanding that:

* Waiver -- Any of the above conditions can be waived if you get
permission from the copyright holder.

* Public Domain -- Where the work or any of its elements is in
the public domain under applicable law, the status is in no way
affected by the license.

* Other Rights -- In no way are any of the following rights
affected by the license:

 - Your fair dealing or fair use rights, or other applicable
copyright exceptions and limitations;

 - The author's moral rights;

 - Rights other persons may have either in the work itself or
in how the work is used, such as publicity or privacy rights.

* Notice -- For any reuse or distribution, you must make clear
to others the license terms of this work.  The best way to do
this is with a link to this web page: