{
  "id": 56478,
  "title": "Challenge Baseline system released",
  "url": "/competitions/freesound-audio-tagging/discussion/56478",
  "author_name": "",
  "post_date": "2018-05-10T07:41:55.793870600Z",
  "votes": 6,
  "comment_count": 1,
  "views": 0,
  "content": "<p>The baseline system for this challenge is now available at\n<a href=\"https://github.com/DCASE-REPO/dcase2018_baseline/tree/master/task2\">https://github.com/DCASE-REPO/dcase2018_baseline/tree/master/task2</a> </p>\n\n<p>The baseline is a 3-layer CNN that reads log mel spectrogram features. The input features, architecture, and hyperparameters are adaptations of what is used in Google's Sound Understanding team (explored in more detail in <a href=\"https://ai.google/research/pubs/pub45611\">https://ai.google/research/pubs/pub45611</a>).</p>\n\n<p>The baseline system was used to enter the submission titled 'Challenge Baseline: Log Mel Spectrogram features, 3-layer CNN' which achieved an MAP@3 of ~0.7 in the public leaderboard after training for ~5 epochs on the entire training dataset.</p>\n\n<p>This was done without any kind of tuning whatsoever, so please feel free to use the baseline as a starting point for your own experiments. We have mentioned several ideas for improvement in the baseline's README.</p>\n\n<p>Good luck!</p>\n\n<p>Manoj</p>\n\n<p>(on behalf of all the challenge organizers)</p>",
  "messages": [
    {
      "id": "326700",
      "postDate": "05/10/2018 07:41:55",
      "content": "<p>The baseline system for this challenge is now available at\n<a href=\"https://github.com/DCASE-REPO/dcase2018_baseline/tree/master/task2\">https://github.com/DCASE-REPO/dcase2018_baseline/tree/master/task2</a> </p>\n\n<p>The baseline is a 3-layer CNN that reads log mel spectrogram features. The input features, architecture, and hyperparameters are adaptations of what is used in Google's Sound Understanding team (explored in more detail in <a href=\"https://ai.google/research/pubs/pub45611\">https://ai.google/research/pubs/pub45611</a>).</p>\n\n<p>The baseline system was used to enter the submission titled 'Challenge Baseline: Log Mel Spectrogram features, 3-layer CNN' which achieved an MAP@3 of ~0.7 in the public leaderboard after training for ~5 epochs on the entire training dataset.</p>\n\n<p>This was done without any kind of tuning whatsoever, so please feel free to use the baseline as a starting point for your own experiments. We have mentioned several ideas for improvement in the baseline's README.</p>\n\n<p>Good luck!</p>\n\n<p>Manoj</p>\n\n<p>(on behalf of all the challenge organizers)</p>",
      "rawMarkdown": "The baseline system for this challenge is now available at\nhttps://github.com/DCASE-REPO/dcase2018_baseline/tree/master/task2 \n\nThe baseline is a 3-layer CNN that reads log mel spectrogram features. The input features, architecture, and hyperparameters are adaptations of what is used in Google's Sound Understanding team (explored in more detail in https://ai.google/research/pubs/pub45611).\n\nThe baseline system was used to enter the submission titled 'Challenge Baseline: Log Mel Spectrogram features, 3-layer CNN' which achieved an MAP@3 of ~0.7 in the public leaderboard after training for ~5 epochs on the entire training dataset.\n\nThis was done without any kind of tuning whatsoever, so please feel free to use the baseline as a starting point for your own experiments. We have mentioned several ideas for improvement in the baseline's README.\n\nGood luck!\n\nManoj\n\n(on behalf of all the challenge organizers)",
      "votes": null
    },
    {
      "id": "732070",
      "postDate": "01/29/2020 13:12:18",
      "content": "<p>Just tried to run the baseline code and I get: <code>tensorflow.python.framework.errors_impl.InvalidArgumentError: Expect 3 fields but have 5 in record 0\n     [[{{node DecodeCSV}}]]\n     [[{{node IteratorGetNext}}]]</code>. </p>\n\n<p>I suppose it's too late for any kind of support?</p>",
      "rawMarkdown": "Just tried to run the baseline code and I get: `tensorflow.python.framework.errors_impl.InvalidArgumentError: Expect 3 fields but have 5 in record 0\n\t [[{{node DecodeCSV}}]]\n\t [[{{node IteratorGetNext}}]]`. \n\nI suppose it's too late for any kind of support?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 732070,
      "author_name": "danijeldomazet",
      "author_url": "",
      "post_date": "01/29/2020 13:12:18",
      "content": "<p>Just tried to run the baseline code and I get: <code>tensorflow.python.framework.errors_impl.InvalidArgumentError: Expect 3 fields but have 5 in record 0\n     [[{{node DecodeCSV}}]]\n     [[{{node IteratorGetNext}}]]</code>. </p>\n\n<p>I suppose it's too late for any kind of support?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "326700": "The baseline system for this challenge is now available at\nhttps://github.com/DCASE-REPO/dcase2018_baseline/tree/master/task2 \n\nThe baseline is a 3-layer CNN that reads log mel spectrogram features. The input features, architecture, and hyperparameters are adaptations of what is used in Google's Sound Understanding team (explored in more detail in https://ai.google/research/pubs/pub45611).\n\nThe baseline system was used to enter the submission titled 'Challenge Baseline: Log Mel Spectrogram features, 3-layer CNN' which achieved an MAP@3 of ~0.7 in the public leaderboard after training for ~5 epochs on the entire training dataset.\n\nThis was done without any kind of tuning whatsoever, so please feel free to use the baseline as a starting point for your own experiments. We have mentioned several ideas for improvement in the baseline's README.\n\nGood luck!\n\nManoj\n\n(on behalf of all the challenge organizers)",
    "732070": "Just tried to run the baseline code and I get: `tensorflow.python.framework.errors_impl.InvalidArgumentError: Expect 3 fields but have 5 in record 0\n\t [[{{node DecodeCSV}}]]\n\t [[{{node IteratorGetNext}}]]`. \n\nI suppose it's too late for any kind of support?"
  },
  "source": "meta"
}