{
  "id": 308211,
  "title": "Melspectrogramed Dataset Available !",
  "url": "/competitions/birdclef-2022/discussion/308211",
  "author_name": "",
  "post_date": "2022-02-17T16:16:50.558830800Z",
  "votes": 27,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Hi, I created Melspectrogramed Image Dataset.</p>\n<p>Each images have (224, 313, 3) shape [n_mels, time_steps (5sec), channels] .</p>\n<p>If long audios cut by 5s, short ones add to 5s.</p>\n<p>Longer Images may make score better.</p>\n<p>I wish this codes help audio competition beginners.</p>\n<p>[Dataset] : <a href=\"https://www.kaggle.com/kaerunantoka/birdclef2022-audio-image-dataset\" target=\"_blank\">https://www.kaggle.com/kaerunantoka/birdclef2022-audio-image-dataset</a></p>\n<p>[Notebook] : <a href=\"https://www.kaggle.com/kaerunantoka/birdclef2022-create-image-data-from-audio-data/notebook\" target=\"_blank\">https://www.kaggle.com/kaerunantoka/birdclef2022-create-image-data-from-audio-data/notebook</a></p>",
  "messages": [
    {
      "id": "1694673",
      "postDate": "02/17/2022 16:16:50",
      "content": "<p>Hi, I created Melspectrogramed Image Dataset.</p>\n<p>Each images have (224, 313, 3) shape [n_mels, time_steps (5sec), channels] .</p>\n<p>If long audios cut by 5s, short ones add to 5s.</p>\n<p>Longer Images may make score better.</p>\n<p>I wish this codes help audio competition beginners.</p>\n<p>[Dataset] : <a href=\"https://www.kaggle.com/kaerunantoka/birdclef2022-audio-image-dataset\" target=\"_blank\">https://www.kaggle.com/kaerunantoka/birdclef2022-audio-image-dataset</a></p>\n<p>[Notebook] : <a href=\"https://www.kaggle.com/kaerunantoka/birdclef2022-create-image-data-from-audio-data/notebook\" target=\"_blank\">https://www.kaggle.com/kaerunantoka/birdclef2022-create-image-data-from-audio-data/notebook</a></p>",
      "rawMarkdown": "Hi, I created Melspectrogramed Image Dataset.\n\nEach images have (224, 313, 3) shape [n_mels, time_steps (5sec), channels] .\n\nIf long audios cut by 5s, short ones add to 5s.\n\nLonger Images may make score better.\n\nI wish this codes help audio competition beginners.\n\n\n[Dataset] : https://www.kaggle.com/kaerunantoka/birdclef2022-audio-image-dataset\n\n[Notebook] : https://www.kaggle.com/kaerunantoka/birdclef2022-create-image-data-from-audio-data/notebook",
      "votes": null
    },
    {
      "id": "1705918",
      "postDate": "02/27/2022 01:13:07",
      "content": "<p>Hey <a href=\"https://www.kaggle.com/kaerunantoka\" target=\"_blank\">@kaerunantoka</a> , I could not find out from where the 313 value for time-steps comes from. Where is it calculated or specified?</p>\n<p>I've been playing with your dataset and it is helping a lot, thank you!</p>",
      "rawMarkdown": "Hey @kaerunantoka , I could not find out from where the 313 value for time-steps comes from. Where is it calculated or specified?\n\nI've been playing with your dataset and it is helping a lot, thank you!",
      "votes": null
    },
    {
      "id": "1777794",
      "postDate": "05/04/2022 19:37:41",
      "content": "<p>By default, <code>hop_length=512</code> in <a href=\"https://librosa.org/doc/main/generated/librosa.feature.melspectrogram.html\" target=\"_blank\">librosa.feature.melspectrogram</a>.<br>\n<code>32000 * 5 / 512 = 312.5</code></p>",
      "rawMarkdown": "By default, `hop_length=512` in [librosa.feature.melspectrogram](https://librosa.org/doc/main/generated/librosa.feature.melspectrogram.html).\n`32000 * 5 / 512 = 312.5`",
      "votes": null
    },
    {
      "id": "1790994",
      "postDate": "05/15/2022 14:51:53",
      "content": "<p>nice work, seems I made something similar a while back, which even it is <a href=\"https://www.kaggle.com/code/jirkaborovec/birdclef-convert-spectrograms-noise-reduce\" target=\"_blank\">kernel</a> can be loaded as an input dataset</p>",
      "rawMarkdown": "nice work, seems I made something similar a while back, which even it is [kernel](https://www.kaggle.com/code/jirkaborovec/birdclef-convert-spectrograms-noise-reduce) can be loaded as an input dataset",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1705918,
      "author_name": "hinepo",
      "author_url": "",
      "post_date": "02/27/2022 01:13:07",
      "content": "<p>Hey <a href=\"https://www.kaggle.com/kaerunantoka\" target=\"_blank\">@kaerunantoka</a> , I could not find out from where the 313 value for time-steps comes from. Where is it calculated or specified?</p>\n<p>I've been playing with your dataset and it is helping a lot, thank you!</p>",
      "votes": null,
      "replies": [
        {
          "id": 1777794,
          "author_name": "shinya7y",
          "author_url": "",
          "post_date": "05/04/2022 19:37:41",
          "content": "<p>By default, <code>hop_length=512</code> in <a href=\"https://librosa.org/doc/main/generated/librosa.feature.melspectrogram.html\" target=\"_blank\">librosa.feature.melspectrogram</a>.<br>\n<code>32000 * 5 / 512 = 312.5</code></p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1790994,
      "author_name": "jirkaborovec",
      "author_url": "",
      "post_date": "05/15/2022 14:51:53",
      "content": "<p>nice work, seems I made something similar a while back, which even it is <a href=\"https://www.kaggle.com/code/jirkaborovec/birdclef-convert-spectrograms-noise-reduce\" target=\"_blank\">kernel</a> can be loaded as an input dataset</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1694673": "Hi, I created Melspectrogramed Image Dataset.\n\nEach images have (224, 313, 3) shape [n_mels, time_steps (5sec), channels] .\n\nIf long audios cut by 5s, short ones add to 5s.\n\nLonger Images may make score better.\n\nI wish this codes help audio competition beginners.\n\n\n[Dataset] : https://www.kaggle.com/kaerunantoka/birdclef2022-audio-image-dataset\n\n[Notebook] : https://www.kaggle.com/kaerunantoka/birdclef2022-create-image-data-from-audio-data/notebook",
    "1705918": "Hey @kaerunantoka , I could not find out from where the 313 value for time-steps comes from. Where is it calculated or specified?\n\nI've been playing with your dataset and it is helping a lot, thank you!",
    "1777794": "By default, `hop_length=512` in [librosa.feature.melspectrogram](https://librosa.org/doc/main/generated/librosa.feature.melspectrogram.html).\n`32000 * 5 / 512 = 312.5`",
    "1790994": "nice work, seems I made something similar a while back, which even it is [kernel](https://www.kaggle.com/code/jirkaborovec/birdclef-convert-spectrograms-noise-reduce) can be loaded as an input dataset"
  },
  "source": "meta"
}