# DFL - Bundesliga Data Shootout — Data

来源：https://www.kaggle.com/competitions/dfl-bundesliga-data-shootout/data

本地 Meta Kaggle 快照；不是本次实时页面。

# Dataset Description

## NOTE: DATA IS NO LONGER AVAILABLE FOR THIS COMPLETED COMPETITION. ##

The competition dataset comprises video recordings of nine football games divided into halves. You're challenged to detect three kinds of player events, both the time of occurrence and the type, within these videos. See the [**Event Descriptions**](https://www.kaggle.com/competitions/dfl-bundesliga-data-shootout/overview/event-descriptions) page for a full description of each event type.

This is a [**Code Competition**](https://www.kaggle.com/competitions/dfl-bundesliga-data-shootout/overview/code-requirements) that will run in [**two stages**](https://www.kaggle.com/competitions/dfl-bundesliga-data-shootout/overview/timeline). During the **training stage**, your submission will only be run against test data for the public leaderboard. The test data for the private leaderboard, however, will comprise games occuring after the training period closes, the **forecasting stage**.

## Training Data ##
- **train/** - Folder containing videos to be used as training data, comprising video recordings from eight games. Both halves are included for four of the games, while only one half is included for the other four games.
- **test/** - Folder containing videos to be used as test data. The test data for the public leaderboard comprises video recordings from one full game and four half-games, the other half of each game being in the training set.
- **clips/** - Short clips from ten additional games, provided without event annotations. You may wish to use these clips to help your model generalize to environments not represented in the training data.
- **train.csv** - Event annotations for videos in the `train/` folder.
  - `video_id` - Identifies which video the event occurred in.
  - `event` - The type of event occurrence, one of `challenge`, `play`, or `throwin`. Also present are labels `start` and `end` indicating the **scoring intervals** of the video. See the [**Evaluation**](https://www.kaggle.com/competitions/dfl-bundesliga-data-shootout/overview/evaluation) page for information about scoring intervals. Scoring intervals are not provided for the test set videos.
  - `event_attributes` - Additional descriptive attributes for the event.
  - `time` - The time, in seconds, the event occurred within the video.

## Example Test Data ##

Please note that this is a [**Code Competition**](https://www.kaggle.com/competitions/dfl-bundesliga-data-shootout/overview/code-requirements), which means you will submit a notebook that makes predictions against a **hidden test set**. The test set visible here is only an example to help you author submission code. When you submit your notebook for scoring, this example data will be replaced by the actual test data.

Note in particular that the example test videos are only short clips (~30 sec.), while the true hidden test videos are full halves (~50 min.). The name and number of videos may also be different. To ensure your submission completes successfully, you should make your code robust to such variations. See our [**Code Competition Debugging**](https://www.kaggle.com/code-competition-debugging) page for tips.