{
  "id": 164345,
  "title": "[Public 0.568, PyTorch] ResNeSt Training & Inference Baseline",
  "url": "/competitions/birdsong-recognition/discussion/164345",
  "author_name": "",
  "post_date": "2020-07-05T23:07:02.946789900Z",
  "votes": 50,
  "comment_count": 5,
  "views": 0,
  "content": "<p>Hi, all.</p>\n\n<p>I published two notebooks using resampled data which I shared in <a href=\"https://www.kaggle.com/c/birdsong-recognition/discussion/164197\">this topic</a>.</p>\n\n<ul>\n<li>Training: <a href=\"https://www.kaggle.com/ttahara/training-birdsong-baseline-resnest50-fast\">https://www.kaggle.com/ttahara/training-birdsong-baseline-resnest50-fast</a></li>\n<li>Inference: <a href=\"https://www.kaggle.com/ttahara/inference-birdsong-baseline-resnest50-fast\">https://www.kaggle.com/ttahara/inference-birdsong-baseline-resnest50-fast</a></li>\n</ul>\n\n<p>I think the data is ok from the result (Public 2nd now).\n<br>\nThis competition seems not so lively for several reasons (e.g. heavy preprocessing, submission problems, ...).\nI hope that you use the data and make competition more competitive and exciting.  </p>\n\n<p>Enjoy kaggling :)</p>\n\n<p><br></p>\n\n<h3>Update</h3>\n\n<p>This notebook achieved <strong>0.568</strong> by changing threshold. But we know <strong>this is only a public LB score</strong>. (see <a href=\"https://www.kaggle.com/c/birdsong-recognition/discussion/167263\">this topic</a> for details).</p>",
  "messages": [
    {
      "id": "916672",
      "postDate": "07/05/2020 23:07:02",
      "content": "<p>Hi, all.</p>\n\n<p>I published two notebooks using resampled data which I shared in <a href=\"https://www.kaggle.com/c/birdsong-recognition/discussion/164197\">this topic</a>.</p>\n\n<ul>\n<li>Training: <a href=\"https://www.kaggle.com/ttahara/training-birdsong-baseline-resnest50-fast\">https://www.kaggle.com/ttahara/training-birdsong-baseline-resnest50-fast</a></li>\n<li>Inference: <a href=\"https://www.kaggle.com/ttahara/inference-birdsong-baseline-resnest50-fast\">https://www.kaggle.com/ttahara/inference-birdsong-baseline-resnest50-fast</a></li>\n</ul>\n\n<p>I think the data is ok from the result (Public 2nd now).\n<br>\nThis competition seems not so lively for several reasons (e.g. heavy preprocessing, submission problems, ...).\nI hope that you use the data and make competition more competitive and exciting.  </p>\n\n<p>Enjoy kaggling :)</p>\n\n<p><br></p>\n\n<h3>Update</h3>\n\n<p>This notebook achieved <strong>0.568</strong> by changing threshold. But we know <strong>this is only a public LB score</strong>. (see <a href=\"https://www.kaggle.com/c/birdsong-recognition/discussion/167263\">this topic</a> for details).</p>",
      "rawMarkdown": "Hi, all.\n\nI published two notebooks using resampled data which I shared in [this topic](https://www.kaggle.com/c/birdsong-recognition/discussion/164197).\n\n* Training: https://www.kaggle.com/ttahara/training-birdsong-baseline-resnest50-fast\n* Inference: https://www.kaggle.com/ttahara/inference-birdsong-baseline-resnest50-fast\n\nI think the data is ok from the result (Public 2nd now).\n<br>\nThis competition seems not so lively for several reasons (e.g. heavy preprocessing, submission problems, ...).\nI hope that you use the data and make competition more competitive and exciting.  \n\nEnjoy kaggling :)\n\n<br>\n### Update\nThis notebook achieved **0.568** by changing threshold. But we know **this is only a public LB score**. (see [this topic](https://www.kaggle.com/c/birdsong-recognition/discussion/167263) for details).",
      "votes": null
    },
    {
      "id": "917284",
      "postDate": "07/06/2020 11:37:24",
      "content": "<p>Cool!!!! good jobs!</p>",
      "rawMarkdown": "Cool!!!! good jobs!",
      "votes": null
    },
    {
      "id": "918065",
      "postDate": "07/07/2020 00:02:59",
      "content": "<p>Thanks for share, i have one question, this model can be used to multiclass classification like ResNet? Sorry i'm very novice.</p>",
      "rawMarkdown": "Thanks for share, i have one question, this model can be used to multiclass classification like ResNet? Sorry i'm very novice.",
      "votes": null
    },
    {
      "id": "919425",
      "postDate": "07/07/2020 21:01:35",
      "content": "<p>Yes.\nDifference between them is backbone architecture. So you can use ResNeSt for multi-class classification by changing classifier head.</p>",
      "rawMarkdown": "Yes.\nDifference between them is backbone architecture. So you can use ResNeSt for multi-class classification by changing classifier head.",
      "votes": null
    },
    {
      "id": "935580",
      "postDate": "07/19/2020 13:59:36",
      "content": "<p>Do you compare it with EFFICIENTNET B3-4</p>",
      "rawMarkdown": "Do you compare it with EFFICIENTNET B3-4",
      "votes": null
    },
    {
      "id": "936761",
      "postDate": "07/20/2020 14:01:18",
      "content": "<p>Not luck with EfficientNet B3, so far.</p>",
      "rawMarkdown": "Not luck with EfficientNet B3, so far.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 917284,
      "author_name": "xiaowangiiiii",
      "author_url": "",
      "post_date": "07/06/2020 11:37:24",
      "content": "<p>Cool!!!! good jobs!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 918065,
      "author_name": "hiramcho",
      "author_url": "",
      "post_date": "07/07/2020 00:02:59",
      "content": "<p>Thanks for share, i have one question, this model can be used to multiclass classification like ResNet? Sorry i'm very novice.</p>",
      "votes": null,
      "replies": [
        {
          "id": 919425,
          "author_name": "ttahara",
          "author_url": "",
          "post_date": "07/07/2020 21:01:35",
          "content": "<p>Yes.\nDifference between them is backbone architecture. So you can use ResNeSt for multi-class classification by changing classifier head.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 935580,
          "author_name": "doanquanvietnamca",
          "author_url": "",
          "post_date": "07/19/2020 13:59:36",
          "content": "<p>Do you compare it with EFFICIENTNET B3-4</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 936761,
          "author_name": "ramarlina",
          "author_url": "",
          "post_date": "07/20/2020 14:01:18",
          "content": "<p>Not luck with EfficientNet B3, so far.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "916672": "Hi, all.\n\nI published two notebooks using resampled data which I shared in [this topic](https://www.kaggle.com/c/birdsong-recognition/discussion/164197).\n\n* Training: https://www.kaggle.com/ttahara/training-birdsong-baseline-resnest50-fast\n* Inference: https://www.kaggle.com/ttahara/inference-birdsong-baseline-resnest50-fast\n\nI think the data is ok from the result (Public 2nd now).\n<br>\nThis competition seems not so lively for several reasons (e.g. heavy preprocessing, submission problems, ...).\nI hope that you use the data and make competition more competitive and exciting.  \n\nEnjoy kaggling :)\n\n<br>\n### Update\nThis notebook achieved **0.568** by changing threshold. But we know **this is only a public LB score**. (see [this topic](https://www.kaggle.com/c/birdsong-recognition/discussion/167263) for details).",
    "917284": "Cool!!!! good jobs!",
    "918065": "Thanks for share, i have one question, this model can be used to multiclass classification like ResNet? Sorry i'm very novice.",
    "919425": "Yes.\nDifference between them is backbone architecture. So you can use ResNeSt for multi-class classification by changing classifier head.",
    "935580": "Do you compare it with EFFICIENTNET B3-4",
    "936761": "Not luck with EfficientNet B3, so far."
  },
  "source": "meta"
}