{
  "id": 217158,
  "title": "a Few Tips on mAP",
  "url": "/competitions/hpa-single-cell-image-classification/discussion/217158",
  "author_name": "tito",
  "post_date": "2021-02-05T14:59:24.086000",
  "votes": 32,
  "comment_count": 2,
  "views": 0,
  "content": "<p>It seems that metric have updated from F1 to mAP.</p>\n<p>I made a notebook to explain some tips on mAP.<br>\n<a href=\"https://www.kaggle.com/its7171/a-few-tips-on-map\" target=\"_blank\">https://www.kaggle.com/its7171/a-few-tips-on-map</a><br>\n(data of vinbigdata competition is used in this notebook, but the essence is the same)</p>\n<p>This may be a bit obvious, but if you are unfamiliar with any of the following items, please read on.</p>\n<ol>\n<li><p>The lower the confidence threshold, the higher the mAP we will get.</p></li>\n<li><p>Class dependence is linear.</p></li>\n<li><p>small sample class is equally or more important as big class.</p></li>\n</ol>",
  "messages": [
    {
      "id": 1187589,
      "postDate": "2021-02-05T14:59:24.087Z",
      "content": "<p>It seems that metric have updated from F1 to mAP.</p>\n<p>I made a notebook to explain some tips on mAP.<br>\n<a href=\"https://www.kaggle.com/its7171/a-few-tips-on-map\" target=\"_blank\">https://www.kaggle.com/its7171/a-few-tips-on-map</a><br>\n(data of vinbigdata competition is used in this notebook, but the essence is the same)</p>\n<p>This may be a bit obvious, but if you are unfamiliar with any of the following items, please read on.</p>\n<ol>\n<li><p>The lower the confidence threshold, the higher the mAP we will get.</p></li>\n<li><p>Class dependence is linear.</p></li>\n<li><p>small sample class is equally or more important as big class.</p></li>\n</ol>",
      "rawMarkdown": "It seems that metric have updated from F1 to mAP.\n\nI made a notebook to explain some tips on mAP.\nhttps://www.kaggle.com/its7171/a-few-tips-on-map\n(data of vinbigdata competition is used in this notebook, but the essence is the same)\n\nThis may be a bit obvious, but if you are unfamiliar with any of the following items, please read on.\n\n1. The lower the confidence threshold, the higher the mAP we will get.\n\n2. Class dependence is linear.\n\n3. small sample class is equally or more important as big class.\n\n",
      "votes": 32
    },
    {
      "id": 1187879,
      "postDate": "2021-02-05T18:42:39.590Z",
      "content": "<p>Tito, thank you for yet another awesome piece of shared information!</p>\n<p>Just so you know, the link in the post is broken. Perhaps, the notebook is still private.</p>",
      "rawMarkdown": "Tito, thank you for yet another awesome piece of shared information!\n\nJust so you know, the link in the post is broken. Perhaps, the notebook is still private.",
      "votes": 2,
      "replies": [
        {
          "id": 1188088,
          "postDate": "2021-02-06T00:14:24.663Z",
          "content": "<p>Hi Raman,</p>\n<p>Thank you for letting me know!!<br>\nI fixed it.</p>",
          "rawMarkdown": "Hi Raman,\n\nThank you for letting me know!!\nI fixed it.",
          "votes": 1
        }
      ]
    }
  ],
  "comments": [
    {
      "id": 1187879,
      "author_name": "Raman",
      "author_url": "",
      "post_date": "2021-02-05T18:42:39.590000",
      "content": "<p>Tito, thank you for yet another awesome piece of shared information!</p>\n<p>Just so you know, the link in the post is broken. Perhaps, the notebook is still private.</p>",
      "votes": 2,
      "replies": [
        {
          "id": 1188088,
          "author_name": "tito",
          "author_url": "",
          "post_date": "2021-02-06T00:14:24.663000",
          "content": "<p>Hi Raman,</p>\n<p>Thank you for letting me know!!<br>\nI fixed it.</p>",
          "votes": 1,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1187589": "It seems that metric have updated from F1 to mAP.\n\nI made a notebook to explain some tips on mAP.\nhttps://www.kaggle.com/its7171/a-few-tips-on-map\n(data of vinbigdata competition is used in this notebook, but the essence is the same)\n\nThis may be a bit obvious, but if you are unfamiliar with any of the following items, please read on.\n\n1. The lower the confidence threshold, the higher the mAP we will get.\n\n2. Class dependence is linear.\n\n3. small sample class is equally or more important as big class.\n\n",
    "1187879": "Tito, thank you for yet another awesome piece of shared information!\n\nJust so you know, the link in the post is broken. Perhaps, the notebook is still private."
  }
}