{
  "id": 237434,
  "title": "Data: Mel Spectrograms for \"train short audio\" dir",
  "url": "/competitions/birdclef-2021/discussion/237434",
  "author_name": "ab",
  "post_date": "2021-05-08T18:47:05.802000",
  "votes": 2,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi, </p>\n<p>I have shared the mels of all the audios from \"train short audio\" dir. If you're overwhelmed by slow data-loaders because of loading the audios and preprocessing, you could use this dataset to directly load the mel and not worry about .ogg files at all. </p>\n<p>It significantly reduces time from 2-3sec per batch (32) to ~8 batches per seconds on google colab (num_workers=2). </p>\n<p>More information is in the data link.</p>\n<p>Please check here: <a href=\"https://www.kaggle.com/idwivedi/birdclef-2021-mels\" target=\"_blank\">https://www.kaggle.com/idwivedi/birdclef-2021-mels</a></p>",
  "messages": [
    {
      "id": 1298345,
      "postDate": "2021-05-08T18:47:05.803Z",
      "content": "<p>Hi, </p>\n<p>I have shared the mels of all the audios from \"train short audio\" dir. If you're overwhelmed by slow data-loaders because of loading the audios and preprocessing, you could use this dataset to directly load the mel and not worry about .ogg files at all. </p>\n<p>It significantly reduces time from 2-3sec per batch (32) to ~8 batches per seconds on google colab (num_workers=2). </p>\n<p>More information is in the data link.</p>\n<p>Please check here: <a href=\"https://www.kaggle.com/idwivedi/birdclef-2021-mels\" target=\"_blank\">https://www.kaggle.com/idwivedi/birdclef-2021-mels</a></p>",
      "rawMarkdown": "Hi, \n\nI have shared the mels of all the audios from \"train short audio\" dir. If you're overwhelmed by slow data-loaders because of loading the audios and preprocessing, you could use this dataset to directly load the mel and not worry about .ogg files at all. \n\nIt significantly reduces time from 2-3sec per batch (32) to ~8 batches per seconds on google colab (num_workers=2). \n\nMore information is in the data link.\n\nPlease check here: https://www.kaggle.com/idwivedi/birdclef-2021-mels",
      "votes": 2
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1298345": "Hi, \n\nI have shared the mels of all the audios from \"train short audio\" dir. If you're overwhelmed by slow data-loaders because of loading the audios and preprocessing, you could use this dataset to directly load the mel and not worry about .ogg files at all. \n\nIt significantly reduces time from 2-3sec per batch (32) to ~8 batches per seconds on google colab (num_workers=2). \n\nMore information is in the data link.\n\nPlease check here: https://www.kaggle.com/idwivedi/birdclef-2021-mels"
  }
}