# DCASE2018 Challenge - Task1B Leaderboard — Data

来源：https://www.kaggle.com/competitions/dcase2018-task1b-leaderboard/data

本地 Meta Kaggle 快照；不是本次实时页面。

# Dataset Description

The data for the competition consists of **training dataset**, 10-second audio segments released together with the scene class labels, and **test dataset** which is released without the scene labels. Your task is to predict the scene labels for the audio segments in the test dataset.

# Datasets

## Training dataset

The development dataset provided with DCASE2018 Challenge task 1B is used as training dataset. There are three methods for obtaining the training dataset.

1) Dataset can be downloaded directly

***

### [TUT Urban Acoustic Scenes 2018 Mobile, Development dataset  (11.3 GB) ](https://zenodo.org/record/1228235)

[![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.1228235.svg)](https://doi.org/10.5281/zenodo.1228235)

***

2) Dataset can be downloaded using the provided download script:

`python task1b_dataset_downloader.py -d train`

3) If you are using official [DCASE2018 Challenge baseline system](https://github.com/DCASE-REPO/dcase2018_baseline/tree/master/task1), you can use it to download the dataset too:

`python task1b.py -m dev --download_dataset datasets/`

### Organization of the dataset

The dataset package contains readme file, which explains how the dataset is organized. Reference data for the provided audio segments can be found in CSV-formated file `meta.csv`:

<pre>filename	scene_label	identifier	source_label
audio/airport-barcelona-0-0-a.wav	airport	barcelona-0	a
audio/airport-barcelona-0-1-a.wav	airport	barcelona-0	a
....
audio/bus-barcelona-17-706-a.wav	bus	barcelona-17	a
audio/bus-helsinki-18-707-a.wav	bus	helsinki-18	a
audio/bus-helsinki-18-708-a.wav	bus	helsinki-18	a
...
audio/metro-london-46-1402-b.wav	metro	london-46	b
audio/metro-stockholm-57-1699-b.wav	metro	stockholm-57	b
...
audio/street_traffic-vienna-178-5490-c.wav	street_traffic	vienna-178	c
audio/tram-helsinki-182-5639-c.wav	tram	helsinki-182	c
...
audio/tram-vienna-285-8637-a.wav	tram	vienna-285	a
audio/tram-vienna-285-8638-a.wav	tram	vienna-285	a
audio/tram-vienna-285-8639-a.wav	tram	vienna-285	a
</pre>

Naming convention for audio segments is <pre>file naming convention: [scene label]-[city]-[location id]-[segment id]-[device id].wav</pre>

`source_label` field in the meta data or `[device id]` field in the audio segment filename is used to denote the used recording device. 

The dataset is provided with single fold cross-validation set (training / testing) to ease the usage of the development dataset. Participants can do their own cross-validation setup too. File lists assigning segments to the subsets are provided under directory `evaluation_setup`. 

## Testing dataset

Similarly to the development dataset, there are three methods for obtaining the dataset.

1) Downloading it directly:

***

### [TUT Urban Acoustic Scenes 2018 Mobile, Leaderboard dataset  (2.3 GB) ](https://zenodo.org/record/1245184)

[![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.1245184.svg)](https://doi.org/10.5281/zenodo.1245184)

***

2) Dataset can be downloaded using the provided download script:

`python task1b_dataset_downloader.py -d test`

3) If you are using official [DCASE2018 Challenge baseline system](https://github.com/DCASE-REPO/dcase2018_baseline/tree/master/task1) you can use it to download the dataset too:

`python task1b.py -m leaderboard --download_dataset datasets/`

## Dataset license

<pre>
    Copyright (c) 2018 Tampere University of Technology and its licensors 
    All rights reserved.
    Permission is hereby granted, without written agreement and without license or royalty 
    fees, to use and copy the TUT Urban Acoustic Scenes 2018 Mobile (“Work”) described in this document 
    and composed of audio and metadata. This grant is only for experimental and non-commercial 
    purposes, provided that the copyright notice in its entirety appear in all copies of this Work, 
    and the original source of this Work, (Audio Research Group from Laboratory of Signal
    Processing at Tampere University of Technology), 
    is acknowledged in any publication that reports research using this Work.
    Any commercial use of the Work or any part thereof is strictly prohibited. 
    Commercial use include, but is not limited to:
    - selling or reproducing the Work
    - selling or distributing the results or content achieved by use of the Work
    - providing services by using the Work. 
    
    IN NO EVENT SHALL TAMPERE UNIVERSITY OF TECHNOLOGY OR ITS LICENSORS BE LIABLE TO ANY PARTY
    FOR DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE
    OF THIS WORK AND ITS DOCUMENTATION, EVEN IF TAMPERE UNIVERSITY OF TECHNOLOGY OR ITS
    LICENSORS HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
    
    TAMPERE UNIVERSITY OF TECHNOLOGY AND ALL ITS LICENSORS SPECIFICALLY DISCLAIMS ANY
    WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND 
    FITNESS FOR A PARTICULAR PURPOSE. THE WORK PROVIDED HEREUNDER IS ON AN "AS IS" BASIS, AND 
    THE TAMPERE UNIVERSITY OF TECHNOLOGY HAS NO OBLIGATION TO PROVIDE MAINTENANCE, SUPPORT, 
    UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
</pre>