{
  "id": 68256,
  "title": "Why Leaderboard highest score is still less than 30 percent?",
  "url": "/competitions/rsna-pneumonia-detection-challenge/discussion/68256",
  "author_name": "Sajratul Yakin Rubaiat",
  "post_date": "2018-10-10T19:42:48.158000",
  "votes": 1,
  "comment_count": 4,
  "views": 0,
  "content": "<p>It's not been so great to say, someone, hey I'm 25% percent sure that you got lung Opacity here amd here. One cause may be the boundary axis(X, Y, W, H) is not matching in 100% perception with test set data. Is that right?</p>",
  "messages": [
    {
      "id": 401865,
      "postDate": "2018-10-10T19:52:23.767Z",
      "content": "<blockquote>\n  <p>hey I'm 25% percent sure that you got lung Opacity</p>\n</blockquote>\n\n<p>The metric here is not confidence. It's rather precision for bounding boxes.\nA classifier can be done with more than 90% accuracy I think (some kernels do it). </p>",
      "rawMarkdown": "&gt; hey I'm 25% percent sure that you got lung Opacity\n\nThe metric here is not confidence. It's rather precision for bounding boxes.\nA classifier can be done with more than 90% accuracy I think (some kernels do it). ",
      "votes": 1,
      "replies": [
        {
          "id": 402000,
          "postDate": "2018-10-11T03:00:16.077Z",
          "content": "<p>May I know which kernel has 90% classification accuracy ? I've been trying to use Chexnet with no luck</p>",
          "rawMarkdown": "May I know which kernel has 90% classification accuracy ? I've been trying to use Chexnet with no luck"
        },
        {
          "id": 402226,
          "postDate": "2018-10-11T11:27:43.060Z",
          "content": "<p>Sorry, 90% is just my guess. I saw only 80+ % in kernels. </p>",
          "rawMarkdown": "Sorry, 90% is just my guess. I saw only 80+ % in kernels. \n"
        },
        {
          "id": 402787,
          "postDate": "2018-10-12T10:15:22.260Z",
          "rawMarkdown": "",
          "votes": 1,
          "isDeleted": true
        }
      ]
    },
    {
      "id": 401861,
      "postDate": "2018-10-10T19:42:48.160Z",
      "content": "<p>It's not been so great to say, someone, hey I'm 25% percent sure that you got lung Opacity here amd here. One cause may be the boundary axis(X, Y, W, H) is not matching in 100% perception with test set data. Is that right?</p>",
      "rawMarkdown": "It's not been so great to say, someone, hey I'm 25% percent sure that you got lung Opacity here amd here. One cause may be the boundary axis(X, Y, W, H) is not matching in 100% perception with test set data. Is that right?",
      "votes": 1
    }
  ],
  "comments": [
    {
      "id": 401865,
      "author_name": "Sergey Zlobin",
      "author_url": "",
      "post_date": "2018-10-10T19:52:23.767000",
      "content": "<blockquote>\n  <p>hey I'm 25% percent sure that you got lung Opacity</p>\n</blockquote>\n\n<p>The metric here is not confidence. It's rather precision for bounding boxes.\nA classifier can be done with more than 90% accuracy I think (some kernels do it). </p>",
      "votes": 1,
      "replies": [
        {
          "id": 402000,
          "author_name": "yukiya",
          "author_url": "",
          "post_date": "2018-10-11T03:00:16.077000",
          "content": "<p>May I know which kernel has 90% classification accuracy ? I've been trying to use Chexnet with no luck</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 402226,
          "author_name": "Sergey Zlobin",
          "author_url": "",
          "post_date": "2018-10-11T11:27:43.060000",
          "content": "<p>Sorry, 90% is just my guess. I saw only 80+ % in kernels. </p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 402787,
          "author_name": "",
          "author_url": "",
          "post_date": "2018-10-12T10:15:22.260000",
          "content": "",
          "votes": 1,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "401865": "&gt; hey I'm 25% percent sure that you got lung Opacity\n\nThe metric here is not confidence. It's rather precision for bounding boxes.\nA classifier can be done with more than 90% accuracy I think (some kernels do it). ",
    "401861": "It's not been so great to say, someone, hey I'm 25% percent sure that you got lung Opacity here amd here. One cause may be the boundary axis(X, Y, W, H) is not matching in 100% perception with test set data. Is that right?"
  }
}