{
  "id": 203394,
  "title": "Datasets for colab users",
  "url": "/competitions/rfcx-species-audio-detection/discussion/203394",
  "author_name": "",
  "post_date": "2020-12-15T04:59:47.083945200Z",
  "votes": 6,
  "comment_count": 2,
  "views": 0,
  "content": "<p>create your own dataset using this notebook: <a href=\"https://www.kaggle.com/gopidurgaprasad/rfcsad-create-dataset-48k\" target=\"_blank\">link</a></p>\n<p>16k dataset for colab users: <a href=\"https://www.kaggle.com/gopidurgaprasad/rfcsad-wav-16k-v1\" target=\"_blank\">link</a></p>\n<p>48k dataset for colab-pro users: <a href=\"https://www.kaggle.com/gopidurgaprasad/rfcsad-wav-48k-v1\" target=\"_blank\">link</a></p>",
  "messages": [
    {
      "id": "1112993",
      "postDate": "12/15/2020 04:59:47",
      "content": "<p>create your own dataset using this notebook: <a href=\"https://www.kaggle.com/gopidurgaprasad/rfcsad-create-dataset-48k\" target=\"_blank\">link</a></p>\n<p>16k dataset for colab users: <a href=\"https://www.kaggle.com/gopidurgaprasad/rfcsad-wav-16k-v1\" target=\"_blank\">link</a></p>\n<p>48k dataset for colab-pro users: <a href=\"https://www.kaggle.com/gopidurgaprasad/rfcsad-wav-48k-v1\" target=\"_blank\">link</a></p>",
      "rawMarkdown": "create your own dataset using this notebook: [link](https://www.kaggle.com/gopidurgaprasad/rfcsad-create-dataset-48k)\n\n16k dataset for colab users: [link](https://www.kaggle.com/gopidurgaprasad/rfcsad-wav-16k-v1)\n\n48k dataset for colab-pro users: [link](https://www.kaggle.com/gopidurgaprasad/rfcsad-wav-48k-v1)",
      "votes": null
    },
    {
      "id": "1117207",
      "postDate": "12/17/2020 19:55:41",
      "content": "<p>A noob question: If we can make use of tfrecords in a faster manner, why do we need those audio files to create the same data representations? Thanks in advance.</p>",
      "rawMarkdown": "A noob question: If we can make use of tfrecords in a faster manner, why do we need those audio files to create the same data representations? Thanks in advance.",
      "votes": null
    },
    {
      "id": "1122458",
      "postDate": "12/22/2020 12:59:27",
      "content": "<p>Something to be aware of when downsampling to 16k is the highest frequency that can be represented at this sample rate is 8k (<a href=\"https://en.wikipedia.org/wiki/Nyquist%E2%80%93Shannon_sampling_theorem\" target=\"_blank\">Nyquist–Shannon sampling theorem</a>) but some species species have a very high pitch (to humans) call in the 10-12khz range.</p>",
      "rawMarkdown": "Something to be aware of when downsampling to 16k is the highest frequency that can be represented at this sample rate is 8k ([Nyquist–Shannon sampling theorem] (https://en.wikipedia.org/wiki/Nyquist%E2%80%93Shannon_sampling_theorem)) but some species species have a very high pitch (to humans) call in the 10-12khz range.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1117207,
      "author_name": "mrutyunjaybiswal",
      "author_url": "",
      "post_date": "12/17/2020 19:55:41",
      "content": "<p>A noob question: If we can make use of tfrecords in a faster manner, why do we need those audio files to create the same data representations? Thanks in advance.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1122458,
      "author_name": "jackvial",
      "author_url": "",
      "post_date": "12/22/2020 12:59:27",
      "content": "<p>Something to be aware of when downsampling to 16k is the highest frequency that can be represented at this sample rate is 8k (<a href=\"https://en.wikipedia.org/wiki/Nyquist%E2%80%93Shannon_sampling_theorem\" target=\"_blank\">Nyquist–Shannon sampling theorem</a>) but some species species have a very high pitch (to humans) call in the 10-12khz range.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1112993": "create your own dataset using this notebook: [link](https://www.kaggle.com/gopidurgaprasad/rfcsad-create-dataset-48k)\n\n16k dataset for colab users: [link](https://www.kaggle.com/gopidurgaprasad/rfcsad-wav-16k-v1)\n\n48k dataset for colab-pro users: [link](https://www.kaggle.com/gopidurgaprasad/rfcsad-wav-48k-v1)",
    "1117207": "A noob question: If we can make use of tfrecords in a faster manner, why do we need those audio files to create the same data representations? Thanks in advance.",
    "1122458": "Something to be aware of when downsampling to 16k is the highest frequency that can be represented at this sample rate is 8k ([Nyquist–Shannon sampling theorem] (https://en.wikipedia.org/wiki/Nyquist%E2%80%93Shannon_sampling_theorem)) but some species species have a very high pitch (to humans) call in the 10-12khz range."
  },
  "source": "meta"
}