{
  "id": 312842,
  "title": "Unusual music ",
  "url": "/competitions/kaggle-pog-series-s01e02/discussion/312842",
  "author_name": "",
  "post_date": "2022-03-14T10:58:02.286407100Z",
  "votes": 9,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I was exploring the data, and I found some really interesting datapoints. I found them when I was trying to convert them to melspectogram, I saw some of were just 'nan' numpy arrays. So I wanted to know what could be the reason for this.<br>\nBelow are my findings:<br>\n1) <strong>Very low volume/amplitude</strong><br>\nfiles -&gt; 010449.ogg , 005589.ogg , 004921.ogg<br>\n2) <strong>No music in majority of time</strong><br>\nfiles -&gt; 019511.ogg , 013375.ogg , 024247.ogg<br>\n3) <strong>NO MUSIC AT ALL</strong><br>\nfile -&gt; 024156.ogg<br>\nThere are many ways to handle them, one simple trick is delete each of them from train, or for 2nd case we can cleverly only take the portion where there is music.</p>\n<p>There may be more of these, I have only seen the train data. If you find some more \"unsual\" music, do share!<br>\nThanks</p>",
  "messages": [
    {
      "id": "1722245",
      "postDate": "03/14/2022 10:58:02",
      "content": "<p>I was exploring the data, and I found some really interesting datapoints. I found them when I was trying to convert them to melspectogram, I saw some of were just 'nan' numpy arrays. So I wanted to know what could be the reason for this.<br>\nBelow are my findings:<br>\n1) <strong>Very low volume/amplitude</strong><br>\nfiles -&gt; 010449.ogg , 005589.ogg , 004921.ogg<br>\n2) <strong>No music in majority of time</strong><br>\nfiles -&gt; 019511.ogg , 013375.ogg , 024247.ogg<br>\n3) <strong>NO MUSIC AT ALL</strong><br>\nfile -&gt; 024156.ogg<br>\nThere are many ways to handle them, one simple trick is delete each of them from train, or for 2nd case we can cleverly only take the portion where there is music.</p>\n<p>There may be more of these, I have only seen the train data. If you find some more \"unsual\" music, do share!<br>\nThanks</p>",
      "rawMarkdown": "I was exploring the data, and I found some really interesting datapoints. I found them when I was trying to convert them to melspectogram, I saw some of were just 'nan' numpy arrays. So I wanted to know what could be the reason for this.\nBelow are my findings:\n1) **Very low volume/amplitude**\nfiles -> 010449.ogg , 005589.ogg , 004921.ogg\n2) **No music in majority of time**\nfiles -> 019511.ogg , 013375.ogg , 024247.ogg\n3) **NO MUSIC AT ALL**\nfile -> 024156.ogg\nThere are many ways to handle them, one simple trick is delete each of them from train, or for 2nd case we can cleverly only take the portion where there is music.\n\nThere may be more of these, I have only seen the train data. If you find some more \"unsual\" music, do share!\nThanks",
      "votes": null
    },
    {
      "id": "1722388",
      "postDate": "03/14/2022 13:46:13",
      "content": "<p>Nice findings <a href=\"https://www.kaggle.com/mrinath\" target=\"_blank\">@mrinath</a>, I also found some audio files which have a very small duration. The ones which are \"NO MUSIC AT ALL\", might be from this small duration category. BTW the file names you mentioned are they the only files which are having <code>nan</code> values in them?</p>\n<p>\n<img src=\"https://i.imgur.com/FAAMkl9.png\">\n</p>",
      "rawMarkdown": "Nice findings @mrinath, I also found some audio files which have a very small duration. The ones which are \"NO MUSIC AT ALL\", might be from this small duration category. BTW the file names you mentioned are they the only files which are having `nan` values in them?\n<p align=\"center\">\n<img width=\"600\", src=\"https://i.imgur.com/FAAMkl9.png\">\n</p>",
      "votes": null
    },
    {
      "id": "1722978",
      "postDate": "03/15/2022 02:51:18",
      "content": "<p>yes they are the only ones,</p>",
      "rawMarkdown": "yes they are the only ones,",
      "votes": null
    },
    {
      "id": "1738763",
      "postDate": "03/29/2022 14:05:41",
      "content": "<p>Adding \"017400.ogg\" with minimum duration of 0.51 sec</p>",
      "rawMarkdown": "Adding \"017400.ogg\" with minimum duration of 0.51 sec",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1722388,
      "author_name": "soumya9977",
      "author_url": "",
      "post_date": "03/14/2022 13:46:13",
      "content": "<p>Nice findings <a href=\"https://www.kaggle.com/mrinath\" target=\"_blank\">@mrinath</a>, I also found some audio files which have a very small duration. The ones which are \"NO MUSIC AT ALL\", might be from this small duration category. BTW the file names you mentioned are they the only files which are having <code>nan</code> values in them?</p>\n<p>\n<img src=\"https://i.imgur.com/FAAMkl9.png\">\n</p>",
      "votes": null,
      "replies": [
        {
          "id": 1722978,
          "author_name": "mrinath",
          "author_url": "",
          "post_date": "03/15/2022 02:51:18",
          "content": "<p>yes they are the only ones,</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1738763,
      "author_name": "hongseokoh",
      "author_url": "",
      "post_date": "03/29/2022 14:05:41",
      "content": "<p>Adding \"017400.ogg\" with minimum duration of 0.51 sec</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1722245": "I was exploring the data, and I found some really interesting datapoints. I found them when I was trying to convert them to melspectogram, I saw some of were just 'nan' numpy arrays. So I wanted to know what could be the reason for this.\nBelow are my findings:\n1) **Very low volume/amplitude**\nfiles -> 010449.ogg , 005589.ogg , 004921.ogg\n2) **No music in majority of time**\nfiles -> 019511.ogg , 013375.ogg , 024247.ogg\n3) **NO MUSIC AT ALL**\nfile -> 024156.ogg\nThere are many ways to handle them, one simple trick is delete each of them from train, or for 2nd case we can cleverly only take the portion where there is music.\n\nThere may be more of these, I have only seen the train data. If you find some more \"unsual\" music, do share!\nThanks",
    "1722388": "Nice findings @mrinath, I also found some audio files which have a very small duration. The ones which are \"NO MUSIC AT ALL\", might be from this small duration category. BTW the file names you mentioned are they the only files which are having `nan` values in them?\n<p align=\"center\">\n<img width=\"600\", src=\"https://i.imgur.com/FAAMkl9.png\">\n</p>",
    "1722978": "yes they are the only ones,",
    "1738763": "Adding \"017400.ogg\" with minimum duration of 0.51 sec"
  },
  "source": "meta"
}