{
  "id": 327041,
  "title": "Audio to Spectogram pipeline",
  "url": "/competitions/birdclef-2022/discussion/327041",
  "author_name": "",
  "post_date": "2022-05-25T10:09:59.846758700Z",
  "votes": 3,
  "comment_count": 2,
  "views": 0,
  "content": "<p>First of all, I would like to congratulate all the competiors of the BirdCLEF 2022. Moreover, thanks to the hosts and in special, my teammates for the excelent job. I have learned a lot!</p>\n<p>As the end of the competition, I would like to share one of our pipelines for audio transformation into melspectrogram. </p>\n<p>Along the competition's month, we tried a lot of different augmentations, treatments, transformations, models, etc. The Audio transformation pipeline helped us in the beggining to try different combinations to create the melspectograms.</p>\n<p>It consists of a pipeline class that allow an easy switch of \"transformations\", that are also designed into classes.</p>\n<p>The main components were parametized by a configuration class for easy modification and consist of :</p>\n<ul>\n<li>Basic Audio treatment ( This one was better done in anothe notebook)</li>\n<li>Step of slicing audios into specific segments (  5s for example)</li>\n<li>Audio to Spectogram<ul>\n<li>Melspectogram</li>\n<li>CPEN</li></ul></li>\n<li>Image Augmentation<ul>\n<li>Frequency Mask</li>\n<li>Time Mask </li>\n<li>Mixups</li></ul></li>\n<li>Save </li>\n</ul>\n<p><a href=\"https://www.kaggle.com/code/paulojunqueira/birdclef2022-audio-to-image\" target=\"_blank\">Audio to Melspectogram Notebook</a></p>\n<p>Hope some of these could help someone in the futere!<br>\nThanks to my teammates:<br>\n<a href=\"https://www.kaggle.com/lucasdmr\" target=\"_blank\">@lucasdmr</a> <br>\n<a href=\"https://www.kaggle.com/hinepo\" target=\"_blank\">@hinepo</a> <br>\n<a href=\"https://www.kaggle.com/gabrielvinicius\" target=\"_blank\">@gabrielvinicius</a> <br>\n<a href=\"https://www.kaggle.com/felipemandrade\" target=\"_blank\">@felipemandrade</a></p>",
  "messages": [
    {
      "id": "1800947",
      "postDate": "05/25/2022 10:09:59",
      "content": "<p>First of all, I would like to congratulate all the competiors of the BirdCLEF 2022. Moreover, thanks to the hosts and in special, my teammates for the excelent job. I have learned a lot!</p>\n<p>As the end of the competition, I would like to share one of our pipelines for audio transformation into melspectrogram. </p>\n<p>Along the competition's month, we tried a lot of different augmentations, treatments, transformations, models, etc. The Audio transformation pipeline helped us in the beggining to try different combinations to create the melspectograms.</p>\n<p>It consists of a pipeline class that allow an easy switch of \"transformations\", that are also designed into classes.</p>\n<p>The main components were parametized by a configuration class for easy modification and consist of :</p>\n<ul>\n<li>Basic Audio treatment ( This one was better done in anothe notebook)</li>\n<li>Step of slicing audios into specific segments (  5s for example)</li>\n<li>Audio to Spectogram<ul>\n<li>Melspectogram</li>\n<li>CPEN</li></ul></li>\n<li>Image Augmentation<ul>\n<li>Frequency Mask</li>\n<li>Time Mask </li>\n<li>Mixups</li></ul></li>\n<li>Save </li>\n</ul>\n<p><a href=\"https://www.kaggle.com/code/paulojunqueira/birdclef2022-audio-to-image\" target=\"_blank\">Audio to Melspectogram Notebook</a></p>\n<p>Hope some of these could help someone in the futere!<br>\nThanks to my teammates:<br>\n<a href=\"https://www.kaggle.com/lucasdmr\" target=\"_blank\">@lucasdmr</a> <br>\n<a href=\"https://www.kaggle.com/hinepo\" target=\"_blank\">@hinepo</a> <br>\n<a href=\"https://www.kaggle.com/gabrielvinicius\" target=\"_blank\">@gabrielvinicius</a> <br>\n<a href=\"https://www.kaggle.com/felipemandrade\" target=\"_blank\">@felipemandrade</a></p>",
      "rawMarkdown": "First of all, I would like to congratulate all the competiors of the BirdCLEF 2022. Moreover, thanks to the hosts and in special, my teammates for the excelent job. I have learned a lot!\n\nAs the end of the competition, I would like to share one of our pipelines for audio transformation into melspectrogram. \n\nAlong the competition's month, we tried a lot of different augmentations, treatments, transformations, models, etc. The Audio transformation pipeline helped us in the beggining to try different combinations to create the melspectograms.\n\nIt consists of a pipeline class that allow an easy switch of \"transformations\", that are also designed into classes.\n\nThe main components were parametized by a configuration class for easy modification and consist of :\n- Basic Audio treatment ( This one was better done in anothe notebook)\n- Step of slicing audios into specific segments (  5s for example)\n- Audio to Spectogram\n     - Melspectogram\n     - CPEN\n- Image Augmentation\n - Frequency Mask\n - Time Mask \n - Mixups\n- Save \n\n\n[Audio to Melspectogram Notebook](https://www.kaggle.com/code/paulojunqueira/birdclef2022-audio-to-image)\n\n\nHope some of these could help someone in the futere!\nThanks to my teammates:\n@lucasdmr \n@hinepo \n@gabrielvinicius \n@felipemandrade",
      "votes": null
    },
    {
      "id": "1801039",
      "postDate": "05/25/2022 11:26:57",
      "content": "<p>Congratulations, this pipeline are robust and with clean code !</p>",
      "rawMarkdown": "Congratulations, this pipeline are robust and with clean code !",
      "votes": null
    },
    {
      "id": "1801217",
      "postDate": "05/25/2022 14:29:55",
      "content": "<p>This comp was such a roller-coaster journey… was a lot of effort and great learning opportunity!</p>",
      "rawMarkdown": "This comp was such a roller-coaster journey... was a lot of effort and great learning opportunity!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1801039,
      "author_name": "gabrielvinicius",
      "author_url": "",
      "post_date": "05/25/2022 11:26:57",
      "content": "<p>Congratulations, this pipeline are robust and with clean code !</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1801217,
      "author_name": "hinepo",
      "author_url": "",
      "post_date": "05/25/2022 14:29:55",
      "content": "<p>This comp was such a roller-coaster journey… was a lot of effort and great learning opportunity!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1800947": "First of all, I would like to congratulate all the competiors of the BirdCLEF 2022. Moreover, thanks to the hosts and in special, my teammates for the excelent job. I have learned a lot!\n\nAs the end of the competition, I would like to share one of our pipelines for audio transformation into melspectrogram. \n\nAlong the competition's month, we tried a lot of different augmentations, treatments, transformations, models, etc. The Audio transformation pipeline helped us in the beggining to try different combinations to create the melspectograms.\n\nIt consists of a pipeline class that allow an easy switch of \"transformations\", that are also designed into classes.\n\nThe main components were parametized by a configuration class for easy modification and consist of :\n- Basic Audio treatment ( This one was better done in anothe notebook)\n- Step of slicing audios into specific segments (  5s for example)\n- Audio to Spectogram\n     - Melspectogram\n     - CPEN\n- Image Augmentation\n - Frequency Mask\n - Time Mask \n - Mixups\n- Save \n\n\n[Audio to Melspectogram Notebook](https://www.kaggle.com/code/paulojunqueira/birdclef2022-audio-to-image)\n\n\nHope some of these could help someone in the futere!\nThanks to my teammates:\n@lucasdmr \n@hinepo \n@gabrielvinicius \n@felipemandrade",
    "1801039": "Congratulations, this pipeline are robust and with clean code !",
    "1801217": "This comp was such a roller-coaster journey... was a lot of effort and great learning opportunity!"
  },
  "source": "meta"
}