{
  "id": 394437,
  "title": "Metric Evaluation",
  "url": "/competitions/birdclef-2023/discussion/394437",
  "author_name": "Zuuk",
  "post_date": "2023-03-13T14:27:34.231000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<p>In the example Notebook for every 5 seconds a single bird is identified.<br>\nIs the metric regards every 5 second audio as independent classification?</p>\n<p>I've observed that in the function def frame_audio()<br>\n    framed_audio = tf.signal.frame(audio_array, frame_length, hop_length, pad_end=True)<br>\n                                                                                       &gt;&gt;&gt; Padding is used ^^^^^^^^^^^^^^ <br>\n    return framed_audio</p>\n<p>as a result, the classification is less reliable in the last frame in the audio file.</p>",
  "messages": [
    {
      "id": 2180004,
      "postDate": "2023-03-13T14:27:34.230Z",
      "content": "<p>In the example Notebook for every 5 seconds a single bird is identified.<br>\nIs the metric regards every 5 second audio as independent classification?</p>\n<p>I've observed that in the function def frame_audio()<br>\n    framed_audio = tf.signal.frame(audio_array, frame_length, hop_length, pad_end=True)<br>\n                                                                                       &gt;&gt;&gt; Padding is used ^^^^^^^^^^^^^^ <br>\n    return framed_audio</p>\n<p>as a result, the classification is less reliable in the last frame in the audio file.</p>",
      "rawMarkdown": "In the example Notebook for every 5 seconds a single bird is identified.\nIs the metric regards every 5 second audio as independent classification?\n\nI've observed that in the function def frame_audio()\n    framed_audio = tf.signal.frame(audio_array, frame_length, hop_length, pad_end=True)\n                                                                                       >>> Padding is used ^^^^^^^^^^^^^^ \n    return framed_audio\n\nas a result, the classification is less reliable in the last frame in the audio file.",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "2180004": "In the example Notebook for every 5 seconds a single bird is identified.\nIs the metric regards every 5 second audio as independent classification?\n\nI've observed that in the function def frame_audio()\n    framed_audio = tf.signal.frame(audio_array, frame_length, hop_length, pad_end=True)\n                                                                                       >>> Padding is used ^^^^^^^^^^^^^^ \n    return framed_audio\n\nas a result, the classification is less reliable in the last frame in the audio file."
  }
}