{
  "fetched_at": "2026-10-19T07:45:30Z",
  "source_url": "https://www.kaggle.com/competitions/acoustic-scene-2018/data",
  "pages": [
    {
      "name": "rules",
      "content": "    <h3>One account per participant</h3>\n    <p>You cannot sign up to Kaggle from multiple accounts and therefore you cannot submit from multiple accounts.</p>\n    <h3>No private sharing outside teams</h3>\n    <p>\n        Privately sharing code or data outside of teams is not permitted.\n        It's okay to share code if made available to all participants on the forums.\n    </p>\n    <h3>Team Mergers</h3>\n    <p>Team mergers are not allowed in this competition.</p>\n\n    <h3>Team Limits</h3>\n    <p>Max team size is limited to 4.</p>\n    <h3>Submission Limits</h3>\n    <p>You may submit a maximum of 2 entries per day.</p>\n    <p>You may select up to 2 final submissions for judging.</p>\n    <h3>Competition Timeline</h3>\n    <p>Start Date: <strong>January 5th, 2018</strong></p>\n    <p>Merger Deadline: <strong>None</strong></p>\n    <p>Entry Deadline: <strong>None</strong></p>\n    <p>End Date: <strong>March 1st, 2018 22:00 UTC (Midnight EET)</strong></p>\n<h3>External Data</h3>\n<p>Use of external data <strong>is not allowed</strong> (rules changed 26. Jan. 2018).</p>\n\n<h3>Test Data</h3>\n<ul>\n<li>Participants are <strong>not allowed</strong> to make subjective judgements of the test data, nor to annotate it. </li>\n<li>Test items has to be classified separately. One cannot use information from other test items when predicting the scene label.</li>\n</ul>\n\n<h3>License</h3>\n<p>Additionally, you agree that the use of provided data is limited to this competition and bound by the End User License Agreement (see \"Data\" section of the competition). Essentially, the EULA prohibits any commercial use of the database.</p>",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "data-description",
      "content": "<p>The data for the competition consists of training data together with the class labels and test data without the labels. Your task is to predict the secret labels for the test data.</p>\n\n<h2>File descriptions</h2>\n<p>The dataset consists of 5 files. Moreover, we have added the license for using the data - <strong>EULA.pdf</strong> </p>\nThe first three listed below are required, and the last two are provided for convenience and may be useful for better crossvalidation accuracy.\n<ul>\n<li><strong>X_train.npy</strong> - the training data: a numpy array of dimension 4500x40x501. This means that there are 4500 training samples, each represented by the mel spectrogram of size 40x501. The first axis (of size 40) is the spectrogram bin index and the second (of size 501) is the time axis.</li>\n<li><strong>X_test.npy</strong> - the test data: a numpy array of dimension 1500x40x501. There are 1500 test samples, each of the same size (40x501) as in the training set.</li>\n<li><strong>y_train.csv</strong> - training set labels: a csv file listing the true classes for each of the 4500 training samples in the same order as they appear in the numpy array X_train. The labels are string format (\"car\", \"beach\",...) and there are altogether 15 classes.</li>\n<li><strong>meta_train.csv</strong> - training set labels with an indicator of the original recording: a csv file with three columns: sample ID, source recording, and class label. You may use this file instead of <strong>y_train.csv</strong> in your cross-validation procedure in order not to use clips from same recording in both training and validation. For example, the first lines of the file are the following:\n<pre>id,identifier,scene_label\n0,I0066,cafe/restaurant\n1,I0030,home\n2,I0330,office\n3,I0278,grocery store\n4,I0120,park\n5,I0033,train\n6,I0278,grocery store\n7,I0401,grocery store\n</pre>\nFor example, samples 3 and 6 are from the same recording, while sample 7 is different. Training with sample 3 and testing with sample 6 will give an overly optimistic score in cross validation.\n</li>\n<li><strong>crossvalidation_train.csv</strong> - a proposed train/test split of the training set. The folds are selected such that samples from the same recording are all on either train or test fold. The columns are as follows.\n<pre>id,scene_label,set\n0,cafe/restaurant,train\n1,home,train\n2,office,test\n3,grocery store,train\n4,park,train\n5,train,train</pre>\n</li>\n</ul>\n\n</ul>",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Description",
      "content": "<p>Acoustic scene classification attempts to categorize audio signals into predetermined classes. Thus the problem setting is similar to speech recognition, except that the target classes are different and a lot more heterogeneous. Yet, there are several possible applications for acoustic event recognition; for example, your mobile phone might automatically detect that you are in a meeting and turn the device to silent mode automatically.</p>\n\n<p>In this competition, the data consists of recordings of 15 different <b>contexts</b> (classes) such as <i>beach</i>, <i>home</i>, <i>restaurant</i>, etc. The recordings have been cut to 10 second long segments, of which the organizers have computed the <i>mel-spectra</i> as the input to your algorithm.</p>\n\n<p>The organizers would like to thank the Audio Research Group of Tampere University of Technology (<a href=\"http://www.tut.fi/en/\">TUT</a>) for kindly providing access to the competition data. Parts of the data have been used for annual DCASE competitions; see e.g., <a href=\"http://www.cs.tut.fi/sgn/arg/dcase2017/index\">DCASE2017</a>.</p>\n\n<p>This competition is part of the TUT course <a href=\"http://www.cs.tut.fi/courses/SGN-41007/\">SGN-41007 Pattern Recognition and Machine Learning</a>.</p>\n\n\n<h2>Prizes</h2>\n<p>Each member of <b>the best TUT team</b> will receive a nerd T-shirt.</p>\n<p><b>The three best TUT teams</b> will receive a certificate signed by the SGN lab leader.</p>\n\n\n<h2>Organizers</h2>\n<ul>\n<li>Heikki Huttunen, Associate Professor, Tampere University of Technology, Finland</li>\n<li>Toni Heittola, Researcher, Tampere University of Technology, Finland</li>\n<li>Antti Happonen, Researcher, Tampere University of Technology, Finland</li>\n</ul>\n\n<h2>References</h2>\n<ol>\n<li> Annamaria Mesaros, Toni Heittola, and Tuomas Virtanen. TUT database for acoustic scene classification and sound event detection. In 24th European Signal Processing Conference 2016 (EUSIPCO 2016), pp 1128-1132. Budapest, Hungary, Aug 2016.</li>\n<li>Annamaria Mesaros, Toni Heittola, Aleksandr Diment, Benjamin Elizalde, Ankit Shah, Emmanuel Vincent, Bhiksha Raj, and Tuomas Virtanen. DCASE 2017 challenge setup: tasks, datasets and baseline system. In Proceedings of the Detection and Classification of Acoustic Scenes and Events 2017 Workshop (DCASE2017), pp 85–92. November 2017. </li>\n</ol>\n",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "<p>The participants will be ranked based on the accuracy of their predictions. The evaluation metric for this competition is the prediction accuracy, <i>i.e.,</i>, the proportion of 10-second test recordings for which your algorithm makes the correct prediction.</p>\n\n<h2>Submission Format</h2>\n<p>Submission files should be in csv format and contain two columns: Id and Prediction. The first column of the csv file is the test sample ID (0,1,2,3,...,1499) and the second column is the class name predicted for that sample by your algorithm.</p>\n<p>The file should contain a header and have the following format:</p>\n<pre>Id,Scene_label<br />0,cafe/restaurant<br />1,grocery store<br />2,beach<br />...<br />1499,restaurant</pre>",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}