{
  "id": 88958,
  "title": "Precision can be improved by pre-classification ?",
  "url": "/competitions/imet-2019-fgvc6/discussion/88958",
  "author_name": "",
  "post_date": "2019-04-11T03:38:17.096853300Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>The labels of iMet have two attributes: \n   Attribute1:   culuture( 398/1102 )\n   Attribute2:  tag( 704/1102 )\nSo , do pre-classification can improve precision ? </p>",
  "messages": [
    {
      "id": "513227",
      "postDate": "04/11/2019 03:38:17",
      "content": "<p>The labels of iMet have two attributes: \n   Attribute1:   culuture( 398/1102 )\n   Attribute2:  tag( 704/1102 )\nSo , do pre-classification can improve precision ? </p>",
      "rawMarkdown": "The labels of iMet have two attributes: \n   Attribute1:   culuture( 398/1102 )\n   Attribute2:  tag( 704/1102 )\nSo , do pre-classification can improve precision ?",
      "votes": null
    },
    {
      "id": "516314",
      "postDate": "04/14/2019 01:33:43",
      "content": "<p>Maybe it works.</p>",
      "rawMarkdown": "Maybe it works.",
      "votes": null
    },
    {
      "id": "523468",
      "postDate": "04/26/2019 11:04:57",
      "content": "<p>If we can found boundary between the two attritutes, then train the net as two part, also depart dataset when do recognize.That's pre-classification.</p>",
      "rawMarkdown": "If we can found boundary between the two attritutes, then train the net as two part, also depart dataset when do recognize.That's pre-classification.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 516314,
      "author_name": "",
      "author_url": "",
      "post_date": "04/14/2019 01:33:43",
      "content": "<p>Maybe it works.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 523468,
      "author_name": "icekstry",
      "author_url": "",
      "post_date": "04/26/2019 11:04:57",
      "content": "<p>If we can found boundary between the two attritutes, then train the net as two part, also depart dataset when do recognize.That's pre-classification.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "513227": "The labels of iMet have two attributes: \n   Attribute1:   culuture( 398/1102 )\n   Attribute2:  tag( 704/1102 )\nSo , do pre-classification can improve precision ?",
    "516314": "Maybe it works.",
    "523468": "If we can found boundary between the two attritutes, then train the net as two part, also depart dataset when do recognize.That's pre-classification."
  },
  "source": "meta"
}