{
  "id": 179051,
  "title": "nnAudio for audio/spec-processing",
  "url": "/competitions/birdsong-recognition/discussion/179051",
  "author_name": "",
  "post_date": "2020-09-01T08:21:14.745249500Z",
  "votes": 6,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Here's an interesting solution vs librosa and torchaudio, for audio/spec-processing.<br>\n\"nnAudio reduces the waveforms-to-spectrograms conversion time for 1,770 waveforms (from the MAPS dataset) from 10.64 seconds with librosa to only 0.001 seconds for Short-Time Fourier Transform (STFT), 18.3 seconds to 0.015 seconds for Mel spectrogram, 103.4 seconds to 0.258 for constant-Q transform (CQT)\"</p>\n<p>Haven't tried it yet but maybe someone else has?</p>\n<p><a href=\"https://arxiv.org/abs/1912.12055\" target=\"_blank\">https://arxiv.org/abs/1912.12055</a><br>\n<a href=\"https://github.com/KinWaiCheuk/nnAudio\" target=\"_blank\">https://github.com/KinWaiCheuk/nnAudio</a></p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F3924325%2Fe89b39638b6f97eb325a6d89bb9af70d%2Fspeedv3.png?generation=1598948355666919&amp;alt=media\" alt=\"\"></p>",
  "messages": [
    {
      "id": "993918",
      "postDate": "09/01/2020 08:21:14",
      "content": "<p>Here's an interesting solution vs librosa and torchaudio, for audio/spec-processing.<br>\n\"nnAudio reduces the waveforms-to-spectrograms conversion time for 1,770 waveforms (from the MAPS dataset) from 10.64 seconds with librosa to only 0.001 seconds for Short-Time Fourier Transform (STFT), 18.3 seconds to 0.015 seconds for Mel spectrogram, 103.4 seconds to 0.258 for constant-Q transform (CQT)\"</p>\n<p>Haven't tried it yet but maybe someone else has?</p>\n<p><a href=\"https://arxiv.org/abs/1912.12055\" target=\"_blank\">https://arxiv.org/abs/1912.12055</a><br>\n<a href=\"https://github.com/KinWaiCheuk/nnAudio\" target=\"_blank\">https://github.com/KinWaiCheuk/nnAudio</a></p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F3924325%2Fe89b39638b6f97eb325a6d89bb9af70d%2Fspeedv3.png?generation=1598948355666919&amp;alt=media\" alt=\"\"></p>",
      "rawMarkdown": "Here's an interesting solution vs librosa and torchaudio, for audio/spec-processing.\n\"nnAudio reduces the waveforms-to-spectrograms conversion time for 1,770 waveforms (from the MAPS dataset) from 10.64 seconds with librosa to only 0.001 seconds for Short-Time Fourier Transform (STFT), 18.3 seconds to 0.015 seconds for Mel spectrogram, 103.4 seconds to 0.258 for constant-Q transform (CQT)\"\n\nHaven't tried it yet but maybe someone else has?\n\nhttps://arxiv.org/abs/1912.12055\nhttps://github.com/KinWaiCheuk/nnAudio\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F3924325%2Fe89b39638b6f97eb325a6d89bb9af70d%2Fspeedv3.png?generation=1598948355666919&alt=media)",
      "votes": null
    },
    {
      "id": "994018",
      "postDate": "09/01/2020 09:56:31",
      "content": "<p>Thank you, it was interesting 👍 </p>",
      "rawMarkdown": "Thank you, it was interesting 👍",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 994018,
      "author_name": "",
      "author_url": "",
      "post_date": "09/01/2020 09:56:31",
      "content": "<p>Thank you, it was interesting 👍 </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "993918": "Here's an interesting solution vs librosa and torchaudio, for audio/spec-processing.\n\"nnAudio reduces the waveforms-to-spectrograms conversion time for 1,770 waveforms (from the MAPS dataset) from 10.64 seconds with librosa to only 0.001 seconds for Short-Time Fourier Transform (STFT), 18.3 seconds to 0.015 seconds for Mel spectrogram, 103.4 seconds to 0.258 for constant-Q transform (CQT)\"\n\nHaven't tried it yet but maybe someone else has?\n\nhttps://arxiv.org/abs/1912.12055\nhttps://github.com/KinWaiCheuk/nnAudio\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F3924325%2Fe89b39638b6f97eb325a6d89bb9af70d%2Fspeedv3.png?generation=1598948355666919&alt=media)",
    "994018": "Thank you, it was interesting 👍"
  },
  "source": "meta"
}